Search

Covid-19 is a virus that has spread and become a global pandemic. This virus infected the vital human organ, which is the lungs. Therefore, this research identified Covid-19 and non-covid-19 diseases based on chest CT-Scan images using K-NN and Naïve Bayes classification methods. The system is constructed through pre-processing, segmentation, GLCM-based feature extraction, and dividing the testing and training data with K-fold cross-validation with the value of 5 and 7, then evaluated using Confusion Matrix. The algorithm accuracy value from the K-NN classification model is obtained as 99,6% and Naïve Bayes got the value of 93,5%. In comparison, the K-NN method obtained the highest sensitivity level with a value of 100% and a specificity value of 98.4% for the two methods used. In this test, the K-NN classifier method is more appropriate than the Naïve Bayes method because some features of GLCM are more accommodating to the KNN classifier. [ABSTRACT FROM AUTHOR]

The goal of this research is to improve the accuracy of semiconductor device health prediction by comparing the efficacy of Enhanced CNN with the Support Vector Machine technique. Support vector machines and Novel Enhanced CNN were the two groups that were used. The parameters were a G-Power of 80%, an alpha of 0.05, and a beta of 0.2. Two iterations were conducted for each group, for a total of 40 iterations and a sample size of 1641. Independent T-tests confirmed the significance of the hypothesis at a level of p = 0.003 (p < 0.05). Based on statistical study with SPSS, the Novel Enhanced CNN and the Support Vector Machine Algorithm had accuracy of 97.10% and 95.85%, respectively. These noteworthy accuracy numbers imply that the Enhanced CNN strategy (97.10% accuracy) works better than the Support Vector Machine methodology (95.85% accuracy) for estimating the health of semiconductor devices. [ABSTRACT FROM AUTHOR]

The main goal of the research is to make agricultural yield predictions more accurate through the use of the Naive Bayes (NB) method with a Recurrent Neural Network (RNN). There were 42 samples used in all, with each group consisting of 21 samples. The initial group applied the NB (NB) technique, but the subsequent group employed the RNN technique. The study was planned with a power of statistics of 80% using G-power. The statistically significant levels were set at an alpha of 0.05 and a beta of 0.10. The results show that the NB algorithm outperforms the RNN algorithm in terms of precision, with a rate of 88% compared to 83% for the latter. The average precision detection is within a vary of ±2 standard variations. The sample-independent t-test yielded a significance value of p = 0.010 (p < 0.05), indicating statistical significance. [ABSTRACT FROM AUTHOR]

This study's objective was to evaluate the effectiveness of Logistic Regression and Linear Regression algorithms in enhancing the precision of semiconductor circuit health estimation. Two separate teams were examined: one utilizing Linear Regression and the other employing a novel Logistic Regression approach. The study employed a G-Power value of 80%, an alpha level of 0.05, and a beta level of 0.2. Statistical analysis utilizing an independent samples t-test, with a significant p-value of 0.000 (p < 0.05), confirmed the hypothesis validity based on a random sample of 20 iterations for each group. SPSS analysis revealed an accuracy of 96.17% for the Novel Logistic Regression technique and 90.77% for the Linear Regression method. These accuracy rates were deemed remarkable, indicating notable performance differences between the two algorithms. Consequently, the findings suggest that in predicting semiconductor device health, the Logistic Regression approach outperforms Linear Regression, which achieved an accuracy of 90.77%. [ABSTRACT FROM AUTHOR]

The major goal is to boost accuracy by comparing the Novel Gradient Boosting Regressor algorithm and the K Nearest Neighbours Classifier approach for analysis and prediction of solar power generation. Materials and Methods: To evaluate, compare, and understand the effectiveness of recommended algorithms, a total of sixty samples were gathered for the two study groups' SPSS analysis. Thirty samples from group 2's K Nearest Neighbours Classifier method and thirty samples from group 1's Novel Gradient Boosting Regressor (GBR) algorithm underwent an 80% power G-power analysis. The accuracy rates observed were 78.6% for the K Nearest Neighbours (KNN) Classifier and 93.3% for the Novel Gradient Boosting Regressor. Consequently, it can be concluded that the GBR method outperforms the KNN Classifier algorithm significantly. [ABSTRACT FROM AUTHOR]

To detect forgeries in signature images using a state-of-the-art deep learning Support Vector Machine (SVM) algorithm based on parameters extracted from the data set. 44 samples total are used in the study, which is split into two groups of 22. Group 1 utilizes CNN-xg, whereas Group 2 employs SVM. Colab software specialized for machine learning is used to run the code. According to simulation findings, the CNN-xg Algorithm obtains a greater reliability of 96.82%, while the SVM achieves reliability of 84.80%; both algorithms have the same significance values of 0.0004 (p < 0.05). CNN-xg identifies forged signatures in the provided dataset more correctly than SVM, demonstrating superior performance. [ABSTRACT FROM AUTHOR]

The current study detects aquatic trash on ocean surfaces by utilizing the Novel Alexnet algorithm in comparison with Gradient Boosting technique. Novel AlexNet algorithm with number of iterations of 20 is compared with Gradient Boosting algorithm with a G-Power of 80% and confidence level of 90% using the clincalc software. The dataset is composed of 324 JPG images depicting marine debris and 724 JPG images of the ocean, which have different levels of clarity and may also include images with noise. After performing this research, The Novel AlexNet Algorithm, boasting an accuracy of 93.77%, outperforms the Gradient Boosting Algorithm, which achieves an accuracy of 92.44%. This discrepancy suggests a statistically significant difference between the two groups, with a significance value of p=0.072 (Independent Sample T-test p<0.05). These findings strongly indicate that the Novel AlexNet algorithm excels in predicting aquatic debris with greater accuracy compared to the Gradient Boosting algorithm. [ABSTRACT FROM AUTHOR]

The objective of this study is to compare the performance of the K-Means algorithm with the Logistic Regression (LR) for the diabetes dataset. Diabetes is a serious and life-threatening disease. Since diabetes cannot be cured, its prevention and treatment have become a critical public health issue. Therefore, in this study, a large sample of data were used from the Kaggle Dataset - Diabetes disease Analysis and Prediction, which included 13 clinical features related to diabetes disease. The sample size is 10. Several control variables have also been included in the analysis to control for potential confounding factors and improve the accuracy of our results. The K Mean and Logistic Regression algorithm has been used for prediction Finally, performed two different analyses to compare the performance of the K Means algorithm with the proposed LR algorithm. The results of these analyses shows that the extended LR method outperformed. The mean accuracy for LR was observed to be 76.8%, while the mean accuracy for K-Means clustering was 46.2%, the LR algorithm was found to be better than K-Means algorithm. The findings of this study indicate that machine learning algorithms have the capability to provide precise predictions of diabetes disease. The p-value for this study is 0.001 which is smaller than p=0.05. This study emphasizes the significance of employing LR for predicting diabetes. In this study, the extended LR algorithm produced an improved accuracy than the K-Means algorithm. However, Logistic Regression still required a larger sample size in order to obtain more accurate results. [ABSTRACT FROM AUTHOR]

This study aimed to find a better way to identify and categorize malware. Researchers compared two algorithms: a Novel Gradient Boosting Algorithm and a Recurrent Neural Network (RNN) Algorithm. They used a larger dataset of malware samples (10,000 entries with 35 details about each). To ensure a reliable comparison, they carefully selected a portion of the data (80%) for training and set a high standard for accuracy (95% confidence level). The results were clear: the Novel Gradient Boosting Algorithm significantly outperformed the other one. It achieved an impressive 91.7% accuracy in identifying malware, whereas the Recurrent Neural Network only managed 79.5%. This difference was statistically proven using a special test (p-value less than 0.005). In conclusion, this study suggests that the Novel Gradient Boosting Algorithm is a much stronger tool for malware detection and classification compared to Recurrent Neural Networks. The researchers recommend using this method in future systems designed to fight malware. [ABSTRACT FROM AUTHOR]

Automatic detection of a particular section from a chosen digital image is having a wider application and this task can be achieved using manual, conventional and dee-learning supported segmentation schemes. This work aims to develop a deep-learning segmentation (DS) approach to extract and examine the waterbodies section from the chosen satellite digital image (SDI). The stages of this scheme include; (i) Image resizing to a required level, (ii) Otsu's based thresholding using Firefly algorithm for image enhancement, (iii) VGG-UNet implementation and (iv) Performance examination by comparing segmented section with the binary mask of waterbodies. This work implements the VGG-UNet based segmentation to extract the required section from the chosen SDI and this scheme is implemented on the raw and the thresholded image for the assessment. The experimental result of this study confirms that the proposed method provides a better segmentation result compared to UNet, UNet+ and UNet++ and this verify the merit of the proposed research. [ABSTRACT FROM AUTHOR]

This study's main goal is to employ a unique Single Shot Multibox identification (SSD) algorithm to predict item identification with a greater rate of accuracy than the Histogram of Oriented Gradient (HOG) approach. Supplies and Procedures: To enhance object detection prediction, a sample of twenty participants was split into two groups of 10 individuals each. The calculation made use of a 95% confidence level, an alpha and beta proportion of 0.05 and 0.2, and a G worth of 0.8. Similar numbers of information tests (N=10) were subjected to both the SSD and Hoard approaches; SSD provided a greater level of accuracy. Findings: The suggested SSD computation has a success rate of 93.47%, whereas the Hoard classifier's rate is 90.43%. The independent sample T-test results in the p-value 0.000 (P0.05), which is less than the 0.05 significance level. This illustrates the statistical importance of the results. Conclusion: SSD model shows higher accuracy in comparison to the Hoard model, for item recognition predictions. [ABSTRACT FROM AUTHOR]

The proposed research work anticipates the regular workout patterns of users by utilising the data collected from their smartphones. The study indicates that the Random Forest algorithm is more effective in predicting human exercise routines with higher precision than the Linear regression algorithm. A dataset of 3000 entry with the headings 'user','activity','Timestamp', and so on are used to compare, with two groups of 10 sample size for each being studied. Linear Regression recorded an accuracy of 73.67%, while the innovative Random Forest Algorithm gives an precision of 76.51. The difference between these groups is statistically significant at 0.033, which is smaller than 0.05 (p<0.05). It is obvious that these groups differ statistically significantly from one another. Compared to the results of the linear regression, the predictions made by the Random Forest algorithm model are more accurate. [ABSTRACT FROM AUTHOR]

The research work is comparative investigation of two elaborate algorithms entitled a novel Evolutionary based Bat Algorithm (EBA) plus a Bat Algorithm (BA) aimed at reducing the power losses in the radial structure through optimal allocation OA in DSTATCOM. Novel Evolutionary gives insight on the Bat Algorithm its utilization for optimal location with the reduced sizing of DSTATCOM under the radial dissemination scheme. The sample predetermined size stands as Gpower software value defining both groups each of 7 sample with total of 14 samples. The G power is estimation for power value of 0.8. Achieved power restoring value of EBA yields regulation power comparative too BA technique. The independent sample t test value 0.003 (p<0.05) stands significant. EBA-based optimization restored power loss reduction (real power loss: 151.9855 KW; reactive power loss: 70.8874 KVAR) than BA-based technique (real power loss: 152.7152 KW; reactive power loss: 71.1048 KVAR). [ABSTRACT FROM AUTHOR]

The aim of this study is an analysis on obesity levels prediction based on smoking habits using stepwise linear regression algorithm in comparison with random forest classifier for improved accuracy. Estimating the incidence of obesity in people using information acquired from the open-access website Kaggle Insufficient weight, normal weight, overweight levels I and II, overweight levels I and II, and obesity types I, II, and III are the categories that can be used to group the 2111 records of the data that have 17 qualities. A comparison between the random forest algorithm (Group 2) and the stepwise linear regression algorithm (Group 1) using 20 records each. The Gpower is 80 % (The values for g power are alpha(α)=0.05 and power=0.85). The research uses the stepwise linear regression algorithm and obtained accuracy of 84.8% while random forest classifier got 81.9 %. The significance value is found to be p=0.001 (p<0.05) after analyzing the results from Independent. This perspective compares the random forest classifier and stepwise linear regression algorithm for predicting obesity levels based on smoking habits. In comparison the stepwise linear regression method performs more accurately than a random forest classifier. [ABSTRACT FROM AUTHOR]

The objective is to determine the improvement in the quality of speech recorded in noisy environments by comparing the Empirical Wavelet Transform with the Discrete Wavelet Transform. The data utilized in this experiment was obtained from the open access Kaggle dataset. To determine the appropriate sample size for our analysis, we used ClinCalc with an alpha value of 0.05, beta of 0.2, and g power of 0.8. The resulting sample size for each group was 45, for a total of 90 participants. To ensure the correctness of our analysis, we separated our dataset into three subsets: 80% for training, 10% for testing, and 10% for validation. Using the SPSS software, we conducted a performance analysis to compare the effectiveness of various techniques. After conducting a thorough analysis, the Empirical wavelet transform algorithm was found to have accuracy of 82%, while another wavelet algorithm on Discrete transform had an accuracy of 65%. An independent samples T-Test was then performed, and the resulting significance value was p=0.001 (p<0.05), demonstrating the statistical relationship between the two strategies. In this research article the Empirical wavelet transform algorithm was found to have a higher accuracy (82%) than the Discrete wavelet transform algorithm (55%). [ABSTRACT FROM AUTHOR]

The main objective of this work is to perform Cirrhosis Stage using Novel CatBoost Classifier Algorithm and Compare it with the Adaboost Classifier Algorithm to improve the accuracy. It went over the in-depth evaluations of recommended algorithms. Ten testing samples were conducted for each algorithm using the Cirrhosis Dataset, which was downloaded from the Kaggle.com website. For the SPSS study, close to ten samples were gathered in order to evaluate, compare, and determine the effectiveness of recommended algorithms. An alpha of 0.05 and a G power of 0.85 are used by the SPSS software. Ten samples each from the Novel CatBoost algorithm group 1 and the Adaboost algorithm group 2 are gathered in order to assess the predictive accuracy of Cirrhosis. We presented and spoke about the results of combining several feature extraction methods with classifiers. Novel CatBoost was shown to have the highest level of accuracy, scoring 92.1%. Based on the independent sample t-test, there is a statistically significant difference between the two groups (p=0.001; p<0.05). to evaluate the precision of the Cirrhosis stage prediction. To improve the accuracy of the Cirrhosis Stage prediction, increasing forecast precision is a very promising approach. The results of this study show that the Novel CatBoost algorithm performs significantly better than the Adaboost approach. [ABSTRACT FROM AUTHOR]

Determining the research goal is to categorize unsolicited electronic messages via the use of a comparison of the RF Algorithm and the SVM Algorithm for purifying junk messages. To determine accuracy percentage, Random Forest and Support Vector Machine algorithms with sample sizes of 10 were analyzed repeatedly with G power values of 0.8 and 0.8, respectively. When comparing the RF Algorithm's accuracy (88.890%) to the SVM Algorithm's accuracy (77.990%), the results reveal that the former produces better outcomes. The significance value of the two-tailed test is 0.001, and p-value of 0.001 (p<0.05) indicates that the difference between the two algorithms is statistically significant. [ABSTRACT FROM AUTHOR]

The purpose of the study is to evaluate how well two different edge detection techniques differentiate between different guava leaf disease kinds. More specifically, a sample size of 20 is used for each approach to compare the performance of the Laplacian Edge Detection algorithms (LED) and Roberts Edge Detection algorithms (RED). In order to guarantee a 95% confidence interval and a scientific significance level of 0.05, this sample size was calculated. The assessment utilises the Guava Disease dataset, which consists of 2300 photos of guava leaves obtained from a Kaggle repository and includes four different diseases that impact the leaves. The analysis shows that the accuracy of the Laplacian edge identification algorithm is 90.83%, whereas the accuracy of the Robert edge detection approach is 89.6%. Based on statistical analysis, there is a significant difference between the two procedures (independent collection t-test p < 0.05) with an associated p-value of 0.003. Notably, when it comes to guava leaf disease classification, the Laplacian detection of edges technique performs noticeably better than the Roberts method. [ABSTRACT FROM AUTHOR]

The aim of this research is to enhance the accuracy of Blockchain based smart voting systems. Secure Hash Algorithm 384-bit (SHA-384) and Secure Hash Algorithm 256-bit (SHA-256) is executed for Blockchain based smart voting systems. The Gpower test used is about 80% (There are 20 samples total,10 for each group variables: alpha 0.05 and beta 0.2and power=0.85). Secure Hash Algorithm 384-bit (SHA-384) (99.45%) has the increased accuracy over Secure Hash Algorithm 256-bit (SHA-256) (99.01%) with a significance value of p=.03, (p<0.05). The accuracy of Secure Hash Algorithm 384-bit (SHA-384) is better when compared to accuracy of Secure Hash Algorithm 256-bit (SHA-256). [ABSTRACT FROM AUTHOR]

The aim of this project is to apply Fuzzy C-means clustering for detecting the presence of gallbladder stones, and then compare the outcomes with those obtained through the usage of K-means clustering. The dataset utilized in this experiment was obtained from the publicly available Kaggle dataset, which consisted of 5,350 images from 726 patients. For the Novel K-means clustering (group-1) and Novel Fuzzy C-means clustering (group-2), 10 samples were selected and compared. The total number of samples was calculated to be 20By using sample computation, it was discovered that the accuracy rate of Novel Fuzzy C-means clustering was 93.27%, while that of K-means clustering was 94.07%. The pretest power was set at 80%, the alpha at 0.05, and the beta at 0.2. This indicates that there is a statistically significant difference (p<0.05) between the two procedures according to an independent sample t-test. It was therefore established that the Novel K-means clustering algorithm performs better as a result of these findings. demonstrated greater accuracy in comparison to the Novel Fuzzy C-means clustering algorithm. [ABSTRACT FROM AUTHOR]

This study compares seed quality analysis using the Innovative Support Vector Machine with the J48 Decision Tree Algorithm. Two groups were evaluated: Group 1 utilized the Innovative Support Vector Machine, while Group 2 employed the J48 Decision Tree Algorithm. Each group consisted of 20 samples, totaling 40 samples overall. The accuracy assessment revealed that the Innovative Support Vector Machine model outperformed the J48 Decision Tree Algorithm, achieving a higher accuracy of 90.44% with statistical significance at the 0.043 level. Additionally, the Innovative Support Vector Machine-based feature extraction technique achieved an accuracy of 94.63%, significantly surpassing the J48 Decision Tree-based feature extraction system. [ABSTRACT FROM AUTHOR]

In order to accurately estimate customer attrition across a range of businesses, this work aims to apply Novel XGBoost (eXtreme Gradient Boosting) over Decision Tree. The study conducted a performance analysis utilizing the Novel XGBoost comparative Decision Tree method on a large data platform to predict customer attrition as accurately as possible. Each of the two algorithms' groups has ten samples, and the statistical analysis was done using the SPSS software with an alpha value of 0.8 and a beta value of 0.2. The G-Power value of 87% was taken into consideration while predicting the dataset's significance value using the graph value. Compared to Decision Trees (72.208%), XG- Boost Algorithms (90.456%) identify customer attrition with the highest accuracy. After that, an independent sample test was carried out with SPSS, and the results showed that a novel XGBoost is statistically significant compared to the current method. The significance value obtained was 0.0085 (p<0.05). The novel XGboost algorithm for Accurate Prediction of Customer Attrition shows greater accuracy when compared to the Decision Tree Algorithm. [ABSTRACT FROM AUTHOR]

This research aimed to predict flight ticket prices using a new XGBoost optimizer (NXGBRO), aiming for better accuracy than the traditional Ridge regression (RR). We compared the two methods using 40 datasets (20 each) and found that the NXGBRO improved overall system performance. To ensure a reliable setup, we employed the ClinCalc software and set specific parameters like alpha (0.5), G-Power (0.8), and a 95% Confidence Interval (CI). Notably, the NXGBRO attained an imposing 82.7%accuracy compared to Ridge regression's 64.4%. The Independent Samples T-Test (ISTT) is a statistical method used to compare the means of precisely two groups. In our case, it was employed to assess whether there is a significant difference in the performance of the NXGBRO compared to an alternative method. Here's what the results indicate significant difference (p-value = 0.000, p<0.05), confirming the NXGBRO's superiority. In conclusion, this study demonstrates that the NXGBRO not only establishes a connection with the RR but also surpasses it in prediction accuracy. [ABSTRACT FROM AUTHOR]

The primary objective of the study is to compare a Novel Elastic Net approach with the Support Vector Machine Technique to identify narcissistic personality disorder. The g-power value is 80%, datasets from various online sources with continuing reviews a limit of 0.05%, certainty span 95%, mean and standard deviation were iterated ten times to collect data for the unique elastic net approach and support vector machine technology. Elastic Net Algorithm predictability is 86%, whereas Support Vector Machine predictability is 81%. The difference in the predictions is statistically significant at p = 0.001, (p<0.05). This indicates both of the methods differ statistically significant in a meaningful way. When compared to Support vector machine technology, the novel elastic net technique provides improved accuracy. [ABSTRACT FROM AUTHOR]

The focus of this study lies in assessment of Hidden Markov Model in User Authentication and compared over accuracy of Support Vector Machine algorithm to increase the security of client authentication. This research study utilizes a dataset of 4000 tuples, known as the keystroke-dynamics dataset. The research project includes using two algorithms, Novel HMM (N=10) and SVM (N=10). ClinCalc software was used to calculate the iterations size for each group, which was set at 10 with a pretest power value of 0.8 and a significance threshold α=0.05. The independent sample t-test yielded a statistically significant result with a significance level of 0.000 (p<0.05), Indicates a statistically significant distinction between the two algorithms. Based on the results, the novel HMM method (84.76% accuracy) has significantly higher mean accuracy compared to SVM (71.61%). When compared to HMM performs significantly better than the Decision tree. [ABSTRACT FROM AUTHOR]

In contrast to the Linear Support Vector Machine (LSVM) algorithm, the objective about the investigation is to increase the accuracy rate of automatic speaker detection by applying a unique MLP Multilayer Perceptron algorithm. The National Centre for Human Language Technology (NCHLT) database provided the research dataset for this study, which had a G-power of 0.8, alpha and beta values of 0.05 and 0.2, and a 95% confidence interval. Automatic speaker recognition is carried out using MLP and SVM, both of which have the same number of data samples (N=10). MLP, however, has a higher accuracy rate. In comparison to the system, the suggested Multilayer Perceptron gets a success rate of 95.66 percent. higher degree of precision compared to the Linear Support Vector Machine (LSVM). Based on a sample size of twenty (ten from Group 1 and ten from Group 2), the enhanced accuracy rate of automatic speaker recognition was calculated, and the result was an 88.90 percent success rate for the linear SVM classifier. This indicates a noteworthy distinction. The study's significance level was found to be p=0.029. In comparison to Linear Support Vector Machine (LSVM), the proposed Multilayer Perceptron (MLP) model offers a higher level of accuracy for the performance analysis of automatic speaker recognition. [ABSTRACT FROM AUTHOR]

The aim of this research work is to compare the accuracy of Adaboost algorithm with K-Nearest Neighbours for the density-based smart traffic system. The dataset named Smart City Traffic Patterns consists of 360,000 images considered for the application. Two groups were selected for this investigation, and Adaboost and K-Nearest Neighbours were the algorithms employed. With g power setting values of 0.05 and 0.85, the test's average Gpower is roughly 85%. The analysis of the traffic system has been done using Adaboost and K-Nearest Neighbours. It shows that there is a statistical difference between adaboost and K-Nearest Neighbours with p=0.013 (the independent sample T-test p<0.05). The proposed method demonstrated strong performance with a mean accuracy of 93%, exceeding the conventional method's 90% accuracy. When compared to K-Nearest Neighbours for the density-based smart traffic system, Adaboost has higher accuracy. [ABSTRACT FROM AUTHOR]

The objective of the paper is to use the Neural Network Support Vector Machine (NNSVM) algorithm to compare the prediction accuracy, so as to predict and classify Indian Currency using the parameters taken from the currency data set with VGG19. For each of the group 130 samples, thus for two groups a total of 260 samples were collected for this investigation. Group 1 uses Neural Network Support Vector Machine while group 2 uses VGG 19 and according to the workflow that was followed, Neural Network Support Vector Machine code has been implemented based on the imported data set using anaconda software and Jupyter notebook is launched. The simulation results shows that Neural Network Support Vector Machine algorithm has 95.4% prediction accuracy whereas VGG 19 has 80.1% prediction accuracy. This gives a significance value of 0.0026 that is (p<0.05). The research result is that for Counterfeit recognition, Neural Network Support Vector Machine has better performance than VGG 19 when used to find the prediction accuracy. [ABSTRACT FROM AUTHOR]

Stock checking is a labour-intensive operation in a warehouse. Therefore, it is suggested to automate the task using a robot capable of autonomous navigation. Simultaneous Localization and Mapping (SLAM) is commonly used in robots to achieve autonomous navigation. However, the SLAM navigation controller cannot achieve linear locomotion in the narrow pathways between shelves and maintain a consistent distance to the shelf simultaneously, which is required to obtain an accurate scanning result. This paper aims to resolve this limitation. Shelf Following Transition Algorithm is proposed to achieve near-linear locomotion in a narrow pathway. Several experiments are conducted in real applications. Based on the result, the robot achieved near-linear locomotion with a kp value of 0.001325, a kd value of 0.0035 and an entry angle of 90 degrees. It can be concluded that the requirement for the stock-checking robot is achieved. [ABSTRACT FROM AUTHOR]

Machine learning (ML) algorithms are getting unsurpassed exposure as a potential technique for solving and modelling the wear behaviour of polymer matrix composites (PMCs). This paper presents the application of ML algorithms in predicting volume loss of reinforced polytetrafluoroethylene (PTFE) matrix composites. Firstly, the Taguchi L27 was harnessed to generate data set in a regulated way. Then multi linear regression (MLR), support vector regression (SVR), particle swarm optimization (PSO) and Harris Hawk's optimization (HHO) coupled with SVR ML algorithms were developed to accurately predict the volume loss of reinforced PTFE matrix composites. Based on the results achieved, it was found that SVR-HHO ML algorithm predicted the volume loss of reinforced PTFE matrix composites better than the other algorithms with determination coefficient (96 %) and root mean square error of 11 %. The ML algorithms could be used for prediction of volume loss of reinforced PTFE matrix composites and development of new PMCs with specific volume loss resistance. [ABSTRACT FROM AUTHOR]

Power supply is a basic component in robot operation. Particularly with mobile robot, portable batteries are crucial to maintain continuous operation. This has led to the development of an autonomous docking system to overcome the hassle of manually recharging the robot. Presently, there are many well-developed docking systems to achieve independent replenishment of battery power by the robot itself. However, most system aims at achieving success at first attempt. As consideration was not put on the adaptability of the docking system, recharging of mobile robot easily fails upon a slight change in the surrounding environment. This paper thus aims to address this issue through development of a robust docking system that is capable in handling any positional error after a failed docking attempt. In the proposed docking system, a custom object detection model was trained to identify location of charging station and subsequently update the latest coordinates into the mobile robot mapping algorithm. Experiment results illustrate the capability of successfully replenishing battery power with charging station at lateral distance up to 80 centimetres from its initial position. [ABSTRACT FROM AUTHOR]

The agglomerative hierarchical clustering technique entails grouping unstructured or unknown observations into several segments. The outcomes of clustering rely on the clustering algorithm which involves distance and linkage methods. There are two groups of distance which are correlation-based and non-correlation-based distance. In this study, the focus is on correlation-based distance because the distance explores the linear relationship between two observations. Furthermore, the distance is also influenced by the occurrence of outliers in a dataset. Therefore, it is crucial to investigate which clustering algorithm exhibits better performance under two different conditions which are with contaminated and without contaminated data set. Next, a simulation study of 5000 iterations is carried out by using MATLAB programming. The average value of the Cophenetic Correlation Coefficient (CCC) is calculated to measure the statistical performance under different number of variables (p=5, p=10, and p=20), different sample sizes, (n = 100, 200, and 500), and different percentage numbers of outlier (10, 20, and 30 percent). From the analysis, it is shown that Spearman correlation is the outperform correlation-based distance that perform well under normal and contaminated data as it constantly produces high value of CCC in both conditions. [ABSTRACT FROM AUTHOR]

Logic mining is a powerful tool for organizations seeking to derive insights from large datasets. By analyzing data and identifying trends, logic mining can help solve problems, prevent losses, and uncover opportunities. In this study, we introduce an innovative approach that uses 3-Satisfiability logical rules and integrates them into the Hopfield neural network to better understand specific datasets. Our primary objective is to develop a robust statistical method called log-linear analysis, which can extract the most relevant attributes and insights. To accomplish this, we employ the 3-Satisfiability Reverse Analysis Method to extract attributes as logical rules from carefully selected 15 datasets. This method serves as a standalone logic mining paradigm, which we seamlessly integrate with the 3-Satisfiability logic within the Hopfield Neural network. Our proposed method assesses and trains datasets generated by standard algorithms. We then compare the performance of the 3-Satisfiability Reverse Analysis results with existing logic mining models, and our proposed method achieves superior accuracy, sensitivity, and Matthews Correlation Coefficient. [ABSTRACT FROM AUTHOR]

The conjugate gradient (CG) procedure is among the widely studied techniques for modelling problems in the areas of deep learning, machine learning, image restoration, neural networks and many more. This is due to their robust convergence and less memory requirements. Recently, numerous modifications of the CG formulas have been presented in literature. However, many of these modifications are very complex with complicated algorithms, while some formulas do not satisfy the descent property or converge globally for general nonlinear functions. In this study, we define a descent modification of the CG formula for unconstrained optimization and blurred image reconstruction. An interesting feature of our formula is that it possesses the descent properties under some mild assumptions. To demonstrate the efficiency and robustness of the new technique, the study considered a set of unconstrained optimization functions and image restoration models. Results from the experimentation shows that the proposed algorithm outperformed other existing algorithms with similar structures on unconstrained optimization problems and reconstructed blurred images with the best accepted quality. [ABSTRACT FROM AUTHOR]

The ultimate goal of clustering analysis is to group observations into clusters where there is maximum similarity within a cluster and dissimilarity between clusters. Due to its effectiveness and simplicity, the K-mean has been one of the most used methods for grouping multivariate data. However, the K-means algorithm is believed to be sensitive to outliers. This study is concerned with the K-means algorithm's capability for clustering functional data or curves especially when data consists of outliers. The performance is compared with K-median, and Kmeans.fd when used to cluster PM10 daily functional data (curves). The methodology involved first building the curves data set using the b-spline basis expansion technique. Then, three groups of data sets were created for comparison analysis: the full set with contaminated data, the normal group, and the sparse (outliers) group. The RobMah technique was used to identify the outliers (sparse) group. The rand index (RI) is used to evaluate the clustering performance. Using a pre-determined K = 2 cluster size obtained from the NbClust, the results have shown that K-means works best for the full data set with contaminated data (i.e., moderate compact curves). Meanwhile, K-median works best for the outliers' (sparse) group (i.e., low compact curves), and Kmeans.fd works best for normal curves (i.e., high compact curves). The results indicate that the method's performance depends on the degree of curve compactness which is also described by the degree of outlying-ness in the data set. The study also comes to the conclusion that the optimum K = 2 provided by NbClust is insufficient and that a better algorithm for evaluating K cluster size is needed when applying K-means to contaminated functional data. This is because the cluster size K = 2 ignores other hidden significant clusters. With K = 2, the algorithms can only distinguish between the group of normal curves and the outliers with a high magnitude, not a low magnitude. This means that when the data set contains outliers, the hybridized idea of functional outlier trimming, filtering, buffering, or sample-reducing techniques is recommended to improve the K-means algorithm under consideration for clustering functional data. [ABSTRACT FROM AUTHOR]

In the burgeoning realm of Artificial Intelligence (AI), the integrity of Expert Systems remains paramount. This research introduces a novel framework geared towards bolstering the trustworthiness and transparency of AI Expert Systems. Anchored in the dual imperatives of ethical considerations and functional efficiency, the study's primary objective was to devise a robust mechanism that demystifies the decision-making processes within these systems. Our methodology melded a rigorous review of existing systems with iterative development and testing of our proposed framework. Findings indicate that our model not only enhances the interpretability of AI Expert Systems but also bolsters user trust, bridging the gap between complex computations and end-users. The implications are profound, offering the potential for widespread adoption in diverse sectors, ensuring AI decisions are both understandable and reliable. Objective: The primary objective was multifaceted. Firstly, the author sought to address the opacity often inherent in AI Expert Systems, making their decision-making processes more comprehensible to end-users. Concurrently, the author aimed to reinforce the trustworthiness of these systems, ensuring their decisions not only made sense to users but were also rooted in robust and ethical computational practices. Methodology: Authors' approach was twofold. He began with a comprehensive review of existing systems, assessing their transparency levels, trust metrics, and any associated challenges. This review provided insights into prevalent gaps and set the stage for the development phase. Drawing from this analysis, the author embarked on crafting the framework, ensuring it was anchored in principles of ethical AI and user-centric design. To validate the model, the author conducted a series of controlled experiments, comparing the system's outputs with traditional Expert Systems in a variety of simulated real-world scenarios. Main Findings: The results were illuminating. The novel framework consistently outperformed traditional models in terms of transparency. Users, ranging from AI experts to laypersons, reported a significantly better understanding of decision-making processes when interacting with the system. Moreover, trust metrics, evaluated through user surveys and objective criteria like error rates and consistency, indicated a marked improvement in trustworthiness. Notably, the system demonstrated an adeptness at offering clear, concise explanations for its decisions, bridging the chasm between intricate algorithms and human comprehensibility. Implications: The ramifications of the study are wide-ranging. By ushering in heightened levels of transparency and trustworthiness, the framework paves the way for broader adoption of AI Expert Systems across sectors. From healthcare and finance to manufacturing and logistics, there exists the potential for AI decisions to be both interpretable and reliable. This not only ensures ethical and efficient AI operation but also fosters greater user confidence and engagement. In conclusion, as AI continues its trajectory into becoming an integral component of various industries, ensuring its trustworthiness and transparency is paramount. This research offers a beacon in this direction, introducing a framework that could redefine the way we perceive and interact with AI Expert Systems. [ABSTRACT FROM AUTHOR]

Forensic science often uses reconstruction techniques to aid in their investigation for evidence. Typically documents containing important information are teared into fragments either by hand or by a shredder. Therefore, the benefit of reconstruction of such shredded documents should be immediately understood. This paper focuses on text based vertically strip cut typed documents, where the strips presented may only be fragments from the same document with no extra and no missing fragments. Existing literature in this area shall be reviewed and suitable methods will be tested against Dhivehi text documents to confirm their efficacy and accuracy in the Maldivian context. The general algorithm is segregated into four major phases: fragment pre-processing, fragment featurizing, fragment compatibility calculation and fragment arrangement. The pre-processing stage involves de-noising the image and fixing skew correction issues that may occur when physically scanning thin, long strips. The featurizing stage is used to calculate features which may be helpful in identifying or comparing the images such as the number of empty columns to the left / right of the fragment. The fragment compatibility calculation simply generates a matrix of values such that in row, i and col, j the value in the matrix cell i, j is equivalent to the probability that fragment j should be placed to the right of fragment i. The final arrangement step involves analyzing the fragment compatibility matrix and generating a list of clusters that indicate the correct arrangement of the fragments which when combined in the aforementioned arrangement would result in a near-perfect reconstruction of the original document. The results show that the applicability of tested methods: pixel edge comparisons, Nearest Neighbor arrangement, and Minimum Spanning Tree arrangement still hold true with Dhivehi text-based documents. [ABSTRACT FROM AUTHOR]

The proliferation of unmanned aerial vehicles (UAVs) across many sectors is seeing a fast growth trajectory, resulting in heightened congestion inside the airspace. As a result, the need to guarantee flight safety and mitigate the risk of accidents among unmanned aerial vehicles has emerged as a critical concern within the rapidly advancing realm of drone technology. Multi agent reinforcement learning presents a viable methodology for tackling these challenges, since it empowers drones to exhibit enhanced intelligence when operating in intricate surroundings alongside several agents. This article presents an examination of multi-agent reinforcement learning and its utilization in augmenting the safety of unmanned aerial vehicles. In this paper, we provide a pragmatic instantiation of multi-agent reinforcement learning, which encompasses the participation of several agents. The research results presented in this study provide evidence of the algorithm's efficacy in reducing drone collisions in intricate and highly populated settings, resulting in a significant rate of success. [ABSTRACT FROM AUTHOR]

Reverse engineering (RE) is a sophisticated science that involves building numerous tools to assist with various jobs. Modern technologies, the most significant of which is the scanner, have contributed to introducing another notion in this field of study. We must decide whether the object is related to engineering, medicine, or archaeology, as well as whether it has surface deformations or is disposable. This research will demonstrate how to utilize software programs, and Arduino UNO approaches to display digital data produced by a device intended to measure product dimensions on a computer without using sophisticated algorithms. The improvement in the dimensions of the data processed by CloudCompare and MATLAB programs reached 96.07%. [ABSTRACT FROM AUTHOR]

Discontinuity properties largely influence the mechanical behavior of rock mass. Orientation grouping is an important part of rock discontinuity characterization. This paper presents a fast weighted optimal spectral clustering method for orientation grouping of rock discontinuity based on 3D point clouds. The key steps include: (1) point cloud preprocessing with normal vector calculation and sharp point filtering, (2) hemispherical raster point generation based on Fibonacci sequence (3) fast spectral clustering algorithm based on raster point weighting (4) optimal group number selection using CHI validity index. Two benchmark rock slope models are adopted for analysis. The results show that the proposed method can effectively improve the computational efficiency and accuracy of orientation grouping. [ABSTRACT FROM AUTHOR]

Accurate segmentation of medical images is required to make the correct diagnosis. This article provides a comprehensive overview of the various successful strategies for producing exact segmentation. Despite this, assembling certain components in a timely manner can be difficult. The adaptive technique of image segmentation could be used to handle a variety of computer vision challenges. A variety of segmentation approaches are reviewed and analysed in this article. The findings of this study can be used to identify when and how to employ image segmentation techniques, as well as to improve their efficacy and accuracy. Image processing comprises several subfields, some of which include object representation, analysis, and data visualization. Segmenting an image is a critical first step in each of these subfields. An image is segmented so that it can be broken into pieces that can be used to meet the needs of a certain application. Manual segmentation is time-consuming and often unnecessary in ordinary situations. This implies that when automatic algorithms fail, human or semi-automated segmentation methods can be used instead. [ABSTRACT FROM AUTHOR]

Planning a tourism trip requires careful consideration, especially the travelling time, budget, and personal interests. As the number of places to visit increases, planning and prioritizing a trip effectively can be challenging. This problem is known as Tourist Trip Design Problem (TTDP). This study aims to solve TTDP that prioritizes user preference and Points of Interest (POIs) attraction popularity. The integer problem formulation of TTDP consists of a single objective function which is to maximize the benefit without exceeding the budget and allocation time. The benefit is optimizing the number of points of interest (POIs) to visit while satisfying all the constraints. A greedy algorithm is proposed that is able to consider user preference and POIs popularity. To minimize the distance traveled between POIs, the algorithm is improved by incorporating the Clarke-Wright Saving algorithm. The performance of the enhanced algorithm is tested on 37 POIs in Langkawi Island, Malaysia. In addition, the operating hours and touring time of POIs were also considered. The start node is fixed at Jetty Point and the tourist is assumed to spend at least one night there. The enhanced algorithm is coded in Matlab and results shows that a maximum of 6 POIs can be visited in a one-day trip. An insight into the itinerary's carbon footprint is also presented, taking into account the use of medium-sized cars by tourists. [ABSTRACT FROM AUTHOR]

This study introduces the Modified Firefly Algorithm (MFA). This refined version merges the conventional Firefly Algorithm with elements like Constraint Satisfaction Problem (CSP), Hamming Distance (HD), objective function filtration, and positional adjustments. By incorporating the Constraint Satisfaction Problem and Hamming Distance into the foundational Firefly Algorithm mechanism, the algorithm can more efficiently navigate and utilize the search space. The inclusion of objective function filtration and positional modifications further refines the performance of the MFA. The Modified Firefly Algorithm showcases significant enhancements in solution quality, success frequency, and performance metrics compared to other Firefly Algorithm iterations, such as the standard Firefly Algorithm, Firefly Algorithm - Tabu Search (FA-TS), and Firefly Algorithm-Variable Neighborhood Descent (FA-VND). Despite the increased computational time necessitated by these added procedures, the superior solution quality of the MFA validates this trade-off. In tests against other Firefly Algorithm versions, the MFA addresses the scheduling issue. Future studies should minimize computation time durations and investigate incorporating more methods to further elevate the MFA's efficiency. [ABSTRACT FROM AUTHOR]

This paper presents the comparison ofselective harmonic elimination in a cascaded h-bridge multilevel (CHBMLI) inverter using PSO and TLBO optimization algorithms. The inverters are employed in a variety of applications, including high-voltage dc transmission, variable drive systems, and electrical vehicles. The inverter's output includes harmonics. PSO and TLBO algorithm is used in this paper To find out the best switching angles for a CHBMLI selective harmonic elimination. In order to get the appropriate fundamental component and get rid of low-order harmonics, this project solves nonlinear equations. These algorithms can produce the most effective triggering pulses to minimize Overall Harmonic Distortion(THD) and Selected Harmonics. If the harmonics are removed, the inverter's output will be improved. MATLAB is used to implement and compared both optimization techniques usingthe seven-level inverter. [ABSTRACT FROM AUTHOR]

This paper introduced a new technique based on the bio-inspired Salp Swarm Algorithm (SSA) that can be used to challenges involving the optimal allocation of Distribution STATic COMpensator (DSTATCOM) in radial distribution systems (RDS). The suggested technique calculates appropriate DSTATCOM placements and sizes utilizing the newly designed Salp Swarm Algorithm. The suggested technique computes load flow utilizing a quick and efficient backward/forward sweep algorithm. The suggested method's goal function is to decrease overall system power losses. The suggested SSA-based optimization strategy in the RDS was demonstrated using a standard IEEE 33 bus system. According to the simulation findings and comparison of the various instances studied, DSTATCOM allocation leads in considerable power loss mitigation with decent voltage profile improvement. [ABSTRACT FROM AUTHOR]

Nowadays the number of consumer in the distribution system is increasing, the electricity demand and the number of capacitor bank installing is also increasing. Improper placement of capacitor bank leads to decrease the benefits of the system. In this article Marine Predator Algorithm (MPA) is proposed to find the optimal location and size of the capacitor. The objective of this work is to minimize power loss of the radial distribution system. The proposed method is implemented on IEEE 34 bus radial distribution system and the results are compared with DE-PS and PGS method. The simulation results show the effectiveness of the proposed method. [ABSTRACT FROM AUTHOR]

Object detection is one of the most challenging problems in computer vision. Detecting an object from an image or video finds applications in different fields like security, medical, automated vehicle systems, etc. Locating instances from a real-world image requires improved accuracy and low latency. Deep learning (DL) techniques have emerged as a powerful tool for training the system and detecting instances. However, DL-based object detection for real-time applications requires a lot of computational resources. Field Programming Gate Array (FPGA) is found to be more promising to give optimized results for object detection using deep learning. This paper presents a detailed study of various DL-based models using different datasets like MS-COCO, PASCAL VOC, etc for object detection. It also provides good insight for selecting an algorithm for an object detection application with FPGA architecture. The study shows that the advancement in software and hardware technologies enables accurate and efficient detection systems for a wide range of applications. [ABSTRACT FROM AUTHOR]

The process of gleaning important facts and insights from vast amounts of unstructured textual data is called text mining, sometimes referred to as text analytics. In today's data-driven business climate, text analytics is becoming increasingly significant. A key component of text mining is term-weighting, which aids in locating significant keywords or phrases in a text document. Text mining algorithms can more precisely identify and analyze documents by giving these phrases weights, which can result in important insights and knowledge discovery. The study article thoroughly analyses several methods put out in the literature to help computers detect language and text. It covers the mathematical underpinnings of term weighting schemes, categorizes them broadly into supervised and statistical methods, and provides examples from both groups. The Vector-Space Model and its variations, which are at the core of many other techniques covered in the paper, are given special emphasis in this article. The significant research highlights the necessity for machines to effectively understand language and material without engaging in plagiarism or other immoral activities. Overall, the study lays the framework for further research in this field by offering a thorough overview of contemporary language and text recognition methods. [ABSTRACT FROM AUTHOR]

The travelling salesperson problem (TSP) is the most studied problem in optimization and computer science. A more challenging version of the Travelling Salesperson Problem (TSP) is the Multiple Travelling Salesperson Problem (MTSP). The goal is to find the shortest path, which is the shortest distance for each salesperson to travel from the depot to each city and return to depot. It is a NP hard problem and has various applications in routing and scheduling. There are various algorithms to solve MTSP such as exact solution method, bioinspired algorithms like genetic ant colony, evolutionary. In this paper we review exact algorithm and bioinspired algorithm. [ABSTRACT FROM AUTHOR]

Diabetic Retinopathy (DR) is a medical condition that arises when there is leakage of blood and other fluids from the tiny retinal blood vessels. If it is left untreated, eye blindness will occur. Diabetic retinopathy has two stages: early stage or Non-Proliferative DR (NPDR), advanced stage or Proliferative DR (PDR). The proposed system detects diabetic retinopathy using Xception algorithm. It classifies early stage of diabetic retinopathy into mild, moderate and severe. So the four categories of detections are mild, moderate, severe and proliferative stage. Fundus images from the datasets are pre-processed to convert image size to 299 299. The deep convolutional neural network utilizes 70% of the images during training phase and 30% in the testing phase. The overall detection accuracy is 97.94% [ABSTRACT FROM AUTHOR]