Search

Various studies have been carried out to classify English accents using traditional classifiers and modern classifiers. In general, research on voice classification and voice recognition that has been done previously uses the MFCC method as voice feature extraction. The stages in this study began with importing datasets, data preprocessing of datasets, then performing MFCC feature extraction, conducting model training, testing model accuracy and displaying a confusion matrix on model accuracy. After that, an analysis of the classification has been carried out. The overall results of the 10 tests on the test set show the highest accuracy value for feature 17 value of 64.96% in the test results obtained some important information, including; The test results on the MFCC coefficient values of twelve to twenty show overfitting. This is shown in the model training process which repeatedly produces high accuracy but produces low accuracy in the classification testing process. The feature assignment on MFCC shows that the higher the feature value assignment on MFCC causes a very large sound feature dimension. With the large number of features obtained, the MFCC method has a weakness in determining the number of features.

In Classification using Support Vector Machine (SVM), each kernel has parameters that affect the classification accuracy results. This study examines the improvement of SVM performance by selecting parameters using Particle Swarm Optimization (PSO) on credit risk classification, the results of which are compared with SVM with random parameter selection. The classification performance is evaluated by applying the SVM classification to the Credit German benchmark credit data set and the private credit data set which is a credit data set issued from a local bank in North Sumatra. Although it requires a longer execution time to achieve optimal accuracy values, the SVM+PSO combination is quite effective and more systematic than trial and error techniques in finding SVM parameter values, so as to produce better accuracy. In general, the test results show that the RBF kernel is able to produce higher accuracy and f1-scores than linear and polynomial kernels. SVM classification with optimization using PSO can produce better accuracy than classification using SVM without optimization, namely the determination of parameters randomly. Credit data classification accuracy increased to 92.31%.

In the effort to hold up the case spread of COVID-19’s growth rate by implementing health protocols such as the use of masks, supervision is needed especially for the people who have not or still have problems to wearing masks. In this research, the system utilizes the robotic power to identify visitors whether they are wearing masks or not, and automatically distribute masks if the user is detected as not wearing a mask. The user face detection process uses the Haar Cascade Classifier algorithm and SVM (Support Vector Machine) to classify users who wear masks or not. For the user who is detected as not wearing masks, myCobot-Pi with the support of suction pump will distribute masks to users. The use of myCobot-Pi as a raspberry pi based robotic arm allows the application of the system on devices that are minimal in terms of specifications and size. Through trials by taking 41 examples of detection cases, 29 cases were found that managed to detect the correct use of masks. In addition, in this study we use PP sheet plastic protector to replace the packaging of the mask because it can be carried by the suction pump properly.

Fuzzy logic is an extension of traditional reasoning, where x is a member of set A or not, or an x can be a member of set A with a certain degree of membership . The ability of fuzzy models to map fuzzy values is the reason for using fuzzy inference models in various cases that use fuzzy values to produce a clear or definite output. In this research, an analysis of the level of accuracy generated by the Sugeno and Mamdani inference model will be carried out in predicting rainfall at Polonia Station, Medan, North Sumatra. Prediction results will be analyzed for accuracy by comparing the results obtained by Sugeno fuzzy inference models and Mamdani using Mean Absolute Percent Error (MAPE). When compared to the results of the Mean Absolute Percent Error (MAPE) Sugeno fuzzy inference model of 1.33% and mamdani fuzzy inference model of 1.45%, then the accuracy rate for the Sugeno inference model is 100%-1.33% = 98.67% while the Mamdani fuzzy inference model is 100%-1.45 = 98.55%. The end result is that the membership function model used in the Sugeno fuzzy model is more accurate than the Mamdani fuzzy inference model in the test case of rainfall prediction at Polonia station in Medan. North Sumatra. The results of the analysis carried out for the Sugeno and Mamdani fuzzy models are influenced by the accuracy of the input values. Rainfall prediction is an important thing to study, weather conditions in certain areas can be predicted so that it can help people's daily activities, can determine a series of community social activities. For example, information on rainfall and its classification is widely used as a guide for agriculture, tourism and transportation, for example: Cropping Patterns, Harvest Predictions, Shipping and flight schedules

Currently, the implementation of teaching and learning at SMP Negeri 1 Binjai Kwala Begumit was done in the classroom alternately. However, with the current condition of pandemic covid-19, the learning process no longer carried out fully in schools. The school has not been using information technology in the form of e-learning applications in the teaching and learning process. The school has difficulty in recording the existing teaching and learning process: assignments, exams, assessments, and other activities. Therefore the use of e-learning applications is now very much needed. With existing school facilities, such as internet facilities and the ICT teachers, training in developing and implementing e-learning for teaching staff become the best alternative so that learning process can be done properly.

Uncertainty of diverse merchant business actors is very much characteristic nowadays; its variants very much depend on the type of business actors, especially digital businesses. In this study we refer to E-Metrics, where every merchant and biller are netted in user behavior activities under the term Financial Technology “acquisition.” Business actors tend to continue to compete in prioritizing innovation so that a Knowledge Acceleration Estimator (KAE) model is needed as performance to facilitate the acceleration of sustainable business under uncertainty in business opportunities, in challenges of electronic kind, or so-called e-metrics.

Data security is one of the most important aspects of today's information era. Some methods are used to secure important data from hackers. The LSB is a steganographic algorithm that is often used to store data in the last bit. In order to improve the security, we combine steganography with cryptography enables. In this research LSB is modified using the multi-bit LSB model. Modifications are made to the bits of each character, the rotation by a certain amount can randomize the plaintext content before cryptographic algorithm, Vernam is performed. The bit on LSB can be inserted data as much as 1, 2, 3 or 4 - bit information. The calculation results of MSE and PSNR values indicate that the use of 1-bit LSB is superior to that of 2-, 3-, or 4-bit LSB.

Handwritten character recognition is considered a complex problem since one’s handwritten character has its characteristics. Data used for this research was a photo of handwritten or scanned handwritten. In this research, Backpropagation Neural Network (BPNN) was used to recognize handwritten Batak Toba character, wherein preprocessing stage feature extraction was done using Diagonal Based Feature Extraction (DBFE) to obtain feature value. Furthermore, the feature value will be used as an input to BPNN. The total number of data used was190 data, where 114 data was used for the training process and another 76 data was used for testing. From the testing process carried out, the accuracy obtained was 87,19 %.

In data security, a technique is needed to secure data from experiments and attacks by cryptanalysts. The technique of securing data is also known as cryptography. In this research, it has been done to improve data security by combining the Hill Cipher Algorithm with the Caesar Cipher Algorithm. This process can be seen from the tests that have been carried out in this study. The result is in a different and sufficiently improved processing time both in terms of decryption and encryption. By combining the two cryptographic algorithms, it is known that message security is much better. It occurs because the encryption and decryption process of the message is carried out using 2 different keys. Based on the tests, it shows that the time of the encryption process with the decryption process on the $5\times 5$ matrix key is very different from using the $3\times 3$ matrix key, the difference is slightly different in time. It shows that the use of a larger matrix key affects the decryption process is very long and also influenced by other cryptographic algorithm processes. In this study, another cryptographic algorithm is the Caesar Cipher Algorithm.

Affine Cipher is the development of Classic Caesar cipher that multiplies plain text before shifting operation. Affine cipher has two parameters namely key a, and b. Key ‘a’ is multiplies the plain text. It must have an inverse with Great Common Divisor (GCD) from the number of characters. The Problems of Affine Cipher Algorithm is the number of characters available and the least possible of key ‘a’ that can be used, so it is considered to be a classic technique that is easy to hack. Generally, Affine Cipher is used to encrypt the plaintext of character ‘A to Z’ so that the number of keys ‘a’ is only 12 digits from 26 digits (n) is available. Development of affine had increased the key ‘a’ into 128 of 256 digits (n) with 8 bit of ASCII. Digraph Transform is the modification of text encryption. It able to square the value of n. Usually, Text is encrypted by one character. In Digraph Transform, text is encrypted by two characters. For two characters, multiplying integer of the first character by number of characters, then add integer of the second character.

The marsh frogs that lost their habitat will become pests that are very detrimental to the fish cultivators. A technological approach is mandatory to assure the fish ponds to always be free from the frogs' presence by detecting the frog sound as a sign of frog existence in the breeding pond. This research implemented Mel Frequency Cepstral Coefficients- Vector Quantization (MFCC-VQ) algorithm to identify the frog's sound. The captured sound signal features will be extracted using Mel Frequency Cepstral Coefficients (MFCC) algorithm. The extracted sound signal will be identified using Vector Quantization (VQ) algorithm. The result of this research is the further the distance of the microphone from the sound source and the noisier environmental conditions in the detection process, the smaller the accuracy percentage. This study's highest accuracy is 99.6% at a distance of 1 meter and 83.2% at a distance of 11 meters in normal circumstances. Through the test, it can also be concluded that the further distance of the microphone from the source, the smaller the percentages of specificity and sensitivity, and the slower the processing speed, although the difference in speed per meter is not too significant.

Text classification using a simple word representation with a linear classifier often considered as strong baselines to gain the best performances. However, a simple word representation like Bag of Word (BOW) has a deficiency of curse dimensionality, so it is only suitable for small datasets. BOW also needs some dependent pre-processing steps like stopwords-removal and stemming. Therefore, the BOW model cannot be implemented automatically because of the dependency in a specific language. On the other hand, deep neural network classifiers can eliminate the pre-processing prerequisite, but this model not efficient in time processing and need a large dataset for the learning process. It becomes a challenge for language that has limitation resources like Bahasa Indonesia. Another novel approach of text classifier is using the fastText model for text classification. This model can minimize pre-processing dependencies and more efficient in training time processing. However, there hasn't been much observation whether the fastText model outperformed the BOW model for small datasets. This paper aims to compare text classification using the TFIDF model as one of the BOW models with a fastText model for 500 news articles in Bahasa Indonesia. The result of this study showed both models gain an outstanding performance, which is 0.97 F-Score. The TFIDF model needs longer pre-processing stages and requiring more training time. Meanwhile, the fastText model only needs to tune some hyperparameters and get similar performance results to the TFIDF model. Based on this study, we can conclude that the fastText model is efficient text classification.

Given the rapid growth of data and the reduced implementation quality of data mining and pattern extraction techniques, the use of feature reduction has become an important challenge of data mining and pattern recognition. An important goal of data reduction techniques is to make the minimum effort and achieve the maximum efficiency of data selection for the implementation of data mining process. The two primary objectives of feature selection are to minimize the errors of the patterns identified in the reduced subset and minimize the number of features. The majority of available feature selection algorithms adopts a single-objective approach. This is the first paper focused on clustering used as the identifier of unsupervised hidden patterns. It is also focused on the principal component analysis (PCA) to analyze the values of the features. The goals of the new multi-objective feature selection problem are to minimize the coefficient of PCA, maximize the accuracy of k-medoids clustering, and minimize the number of selected features. Another innovation of this study was to select the best subset of features at the best performance by using the electromagnetism-like mechanism (EM) algorithm. The proposed method was tested on 14 standard UCI datasets. The results indicated the competitive advantage of this algorithm over other algorithms implemented to solve this problem.

Cryptography has an important role in the process of data security. One of the benefits of applying cryptography is creating a sense of security in hiding data without having to doubt how much data you have. LUC, El Gamal and RSA have different characters, like formatting key processes, encryption and decryption process. So we need an algorithm performance comparison. Steps taken include determining the public key and private key, encryption process, decryption process, and testing in terms of program implementation time using several different texts. Based on tests of the encryption process, the results show that the fastest algorithm is the RSA algorithm. Other than that, the fastest decryption process is the LUC algorithm.

This study comparing the performance of SAW and MOORA methods are in the decision-making process against the evaluation of teacher assistant performance. Then, applying Rank Order Centroid (ROC) is to overcome the process of weighting attributes in decision-making. Subsequently, the data has used by Teaching Assistant Evaluation obtained from UCI Machine Learning Repository, which has 151 instances, five attributes, and 1 class label with multivariate type dataset. The results of testing the calculation of SAW and MOORA methods, obtain the different results where A134 is the best alternative in the calculation of SAW and A132 is the best alternative in the calculation of MOORA. Afterward, if viewing based on running time, the execution process of the SAW method is 0.06 seconds,while the execution process of the MOORA method is 0.02 seconds. In this research, both methods get the same results in ranking, but base on running time, the MOORA method is faster than the SAW method.

This paper proposes and develops an Applying the fuzzy inference method by Mamdani, Tsukamoto and Sugeno. Author will produce a comparison in the final results for each method inthesame case. And this method will be tested on a design model in the process of determining or predicting the export price of a commodity to obtain an optimal profit from the sales of the export commodity. This model includes the prediction of the resale price of exported goods abroad. The process of determining the selling price is calculated using the Fuzzy fuzzyinference method by Mamdani, Tsukamoto, and Sugeno. The variables to be observed are the demand for goods, availability of goods in the farmers, and market price in the foreign countries. The output is an estimated resale price and the profit of each export commodity and the comparison of the selling prices using Mamdani, Tsukamoto, and Sugeno is 2.1% while Mamdani and Sugeno are only 1%after applying logical negation to the basic rules, it is applied to 1.6%. With this comparison, the writer will evaluate the rule base section of Tsukamoto. The results obtained in this study include the prediction model of the resale price of exported goods and the results of the comparison of the performance of the fuzzy interference byMamdani and Tsukamoto logic methods in optimization to provide a more optimal price.

Digital image processing is used to analyze images consisting of visual perceptions whose purpose is to manipulate, modify the image, improve the quality of the image and prioritize the special features contained in the image by using the special attributes of the image characteristics that can be grouped based on the similarity of the image characteristics with each other. One method used in image recognition from artificial neural networks is Learning Vector Quantization (LVQ). The application of the LVQ method for face recognition requires a feature segmentation process so that the feature extraction values obtained only cover the face, this stage is carried out using the Fuzzy C-Means and K-Means Clustering methods. The results of face recognition testing using the LVQ method and K-Means Clustering segmentation obtained an accuracy of 92.22% (training) and 72.5% (testing). Face recognition with Fuzzy C-Means Clustering segmentation obtained face recognition results of 95.78% (training) and 76% (testing). The LVQ-FCM method is better than the LVQ with KMeans, obtained training accuracy of 3.56% and 3.5% for network testing.

Teacher is a profession that has important role for the progress of education literacy, primarily in this current digitalization era that implements the online learning system. Therefore, the assesment of teacher’s performance during online learning is needed to find the advantages and disadvantages of each teacher, with an aim to get an evaluation that can be utilized to fix or improve the teacher’s performance. This study proposes a decision support system that applies the Vikor method as a solution to get the result of teacher’s performance assessment during online learning and make it easier for the decision makers. By using 4 research criteria and 5 alternatives, this research shows that A5 on behalf of Kayan Marbun with a value of 0.5025 is chosen as the teacher with the best performance.

Pengelolaan perkebunan kelapa sawit sering mengalami kendala, antara lain masalah organisme pengganggu tumbuhan (OPT) terutama masalah penyakit. Oleh karena itu, dibuatlah pendekatan untuk mengenali penyakit pada daun kelapa sawit agar dapat membantu kinerja dari para petani kelapa sawit dalam menentukan jenis penyakit pada daun sehingga mendapatkan hasil yang lebih maksimal. Deteksi tepi adalah perubahan nilai intensitas derajat keabuan yang mendadak (besar) dalam jarak yang singkat. Sobel operator digunakan untuk pengidentifikasikan pola wajah, khususnya terdapat di dalam algoritma deteksi tepi. Support Vector Machine (SVM) digunakan sebagai metode klasifikasi. Oleh karena itu, dalam penelitian ini penulis akan menerapkan metode deteksi tepi dengan menggabungkan teknik algoritma Sobel Operator untuk menghilangkan derau dan metode Support Vector Machine sebagai pengklasifikasian data penyakit pada daun kelapa sawit. The management of oil palm plantations often experiences obstacles, including problems with plant pest organisms (OPT), especially disease problems. Therefore, an approach was made to encourage the disease in the leaves of oil palm so that it can help the performance of oil palm farmers in determining the type of disease in the leaves so as to get maximum results. Edge detection is a change in the value of the sudden intensity of the degree of gray (large) in a short distance. Sobel operators are used to identifying face patterns, especially those found in edge detection algorithms. Support Vector Machine (SVM) is used as a classification method. Therefore in this study, the author will apply the edge detection method by combining the Sobel Operator algorithm technique to eliminate noise and the Support Vector Machine method as a classification of disease data on palm oil leaves.

Population size of classical genetic algorithm is determined constantly. Its size remains constant over the run. For more complex problems, larger population sizes need to be avoided from early convergence to produce local optimum. Objective of this research is to evaluate population resizing i.e. dynamic population sizing for Genetic Algorithm (GA) using cloning strategy. We compare performance of proposed method and traditional GA employed to Travelling Salesman Problem (TSP) of A280.tsp taken from TSPLIB. Result shown that GA with dynamic population size exceed computational time of traditional GA.

Starting from the success of giant web service companies as well as Google and Facebook in managing and utilizing unstructured data in the form of consumer generated media and click stream in a very large volume, a concept known as Big Data then became the center attention in the world of information technology. The fact also shows that more and more organizations in the world, whether private companies or government agencies, have difficulty managing data whose volumes are growing and their types are increasingly complex. They have to organize and analyze these data, and they must find the meaning or value of the ever-expanding and increasingly complex data pile, which is said to have exceeded the capability of conventional data processing applications to process it. The condition of this kind of data is also categorized as Big Data, which is interpreted as a set of data in a very large number of challenges lies in how the data should be stored, how to search in the pile of data, how to distribute it, how to visualize it, and how the data should be analyzed. The long-term goal of IbKIK's proposed program is the establishment of a startup company in the field of analytic data from the world of campus directly. Within the planned three-year period, it is desirable that the company be financially self-sufficient by being a data analytics consultant and also creating a sophisticated and advanced Big Data Analytic application platform product. The advantages possessed when the company started from the academic world is the quantity and quality of human resources as intellectual actors can be selected quickly and accurately. Especially with the synchronization between the curriculum content that is taught with its implementation directly through the program IbKIK become useful products and economic value. From the academic point of view, the desired outcomes are from this program published several journals and proceedings of national and international scale, the publication of textbooks, getting HKI, and publications in the mass media. Also, with the success of this company can produce a derivative company engaged in other areas that are still related as a supporter of the business. The product output of the community service activity that has been done for the first year of the planned three years period is the establishment of a Product Information System Sold on E-Commerce Transactions at Market Place. Also, has been established research unit as the forerunner of the business unit under the auspices of the Faculty of Computer Science and Information Technology University of Sumatera Utara under the name Data Science Research Group.

Proses pembelajaran neural network merupakan hal yang penting, bertujuan untuk mengenali lingkungan. Proses pembelajaran neural network membutuhkan waktu untuk dapat mengenali lingkungan. Terutama pada salah satu algoritma neural network yaitu resilient backpropagation. Proses untuk mempercepat pembelajaran resilient backpropagation pada penelitian ini adalah menggunakan teknik parallel processing. Teknik parallel processing yang digunakan adalah multithreading. Teknik parallel ini diterapkan pada bagian hidden layer yaitu membagi jumlah neuron pada hidden layer menjadi beberapa subproses yang dikerjakan secara bersamaan, pembagian yang dilakukan berdasarkan pada jumlah thread. Hasil yang didapatkan dalam penerapan parallel processing menggunakan teknik multithreading ke dalam algoritma resilient backpropagation membantu mempercepat waktu proses pembelajaran resilient backpropagation dengan thread yang digunakan sebanyak 3 buah thread.

Algoritma backpropagation merupakan multi layer perceptron yang banyak digunakan untuk menyelesaikan persoalan yang luas, namun algoritma backpropagation juga mempunyai keterbatasan yaitu laju konvergensi yang cukup lambat. Pada penelitian ini penulis menambahkan parameter learning rate secara adaptif pada setiap iterasi dan koefisien momentum untuk menghitung proses perubahan bobot. Dari hasil simulasi komputer maka diperoleh perbandingan antara algoritma backpropagation standar dengan backpropagation dengan penambahan momentum. Untuk algoritma backpropagation standar kecepatan konvergensi 727 epoch dengan nilai MSE 0,01, sedangkan algoritma backpropagation standar mencapai 4000 epoch dengan nilai MSE 0,001. . Hal ini menunjukkan bahwa algoritma backpropagation adaptive learning lebih cepat mencapai konvergensi daripada algoritma backpropagation standar.

Conceptually, data science burdened with many other terms that refer to various interests about data. The terms calls the subject of learning and study as scientific fields that categorized as the basis of science, implications, and various applications about concepts, theories, or technologies related to data science. The term has a broad impact and affects the scope of work of data scientists. This paper aims to express some of the interests of the three things from the point of that literature for generating consequences.

Backpropagation (BP) has long training time and slow convergence, i.e. in minimizing Mean Square Error (MSE). Adaptive Moment Estimation (ADAM) was employed to speed up training and to obtain rate of BP. Result shown that the minimum MSE obtained was 0.002479194 and 0.002281315 in of 1000 and 2000, respectively.

In daily life, humans do all activities that are dense, both heavy and light activities. Of course, it will require energy or energy to carry out all these activities. The energy needed by humans can be obtained through the consumption of food and beverages, including food and beverages containing alcohol. Ethanol in alcoholic food and beverages human consumption has been consumed since time immemorial with a variety of different uses. Consumption of large amounts of alcohol will cause someone to be drunk or drunk. In some foods and beverages that are on the market, many contain alcohol, even in small quantities. If alcoholic beverages are consumed continuously in excessive amounts, it can cause acute respiratory failure and death. Because ethanol can eliminate awareness, humans who consume it can do bad deeds that are not realized. Accidents that occur in driving are part of other implications of the abuse of alcohol use. Consumption of alcohol which is too much lowers human performance in driving, so increasing the risk of accident or reduce driver survival chances. About 40% of total deaths that occur in traffic are related to alcohol use. A system for detecting alcohol use is needed by drivers, with the aim of reducing this. This system uses a sensor-based MQ303A microcontroller ATmega328 with GSM module SIM800L. This device uses the programming language C and datalog stored on the server using internet communication.

The ant colony algorithm is an algorithm adopted from the behavior of ants which naturally ants are able to find the shortest route on the way from the nest to places of food sources based on footprints on the track that has been passed. The ant colony algorithm helps a lot in solving several problems such as scheduling, traveling salesman problems (TSP) and vehicle routing problems (VRP). In addition, ant colony has been developed and has several variants. However, in its function to find the shortest distance is optimized by utilizing several normalized distance formulas with the data used in finding distances between merchants in the mercant ecosystem. Where in the test normalized distance is superior Hamming distance in finding the shortest distance of 0.2875, then followed by the same value, namely the normalized formula Manhattan distance and normalized Euclidean distance with a value of 0.4675 and without using the normalized distance formula or the original ant colony algorithm gets a value 0.6635. Given the sensitivity in distance search using merchant ecosystem data, the method works well on the ant colony Algorithm using normalized Hamming distance

Words embedding or distributed representations is a popular method for representing words. In this method, the resulting vector value is a set of real values with specific dimensions that are more effective than the Bag of Word (BoW) method. Also, the advantages of distributed representations can produce word vectors that contain semantic and syntactic information, so that word vectors with close meanings will have close word vectors. However, distributed representation requires a huge corpus with a long training time. For this reason, many researchers have created trained word vectors that can be used repeatedly. The problem is that the available trained word vectors are usually general domain word vectors. This study aims to form pre-trained word vectors for specific domains, namely computers and information technology. Researchers used a dataset of student scientific papers from the Universitas Sumatera Utara (USU) repository. Researchers used the word2vec model, where the model has two architectures, namely the Continuous Bag-of-Words (CBOW) and Skip-gram. This research’s result is word2vec model with the CBOW method is more effective than the Skip-gram method.

A person’s self-development is influenced by his ability in solving problems and adapting to his environment. This ability is commonly known as intelligence. Every person has a different dominant intelligence. Honed intelligence can result in managed self-development and self-mastery. A person can discover his talent and expertise to be more advanced in his field. However, discovering one’s dominant intelligence requires several variables to assess his behaviours. This research makes use of a person’s activities on the social media Twitter. Twitter is a platform on which its users share thoughts. Posted tweets are used as objects to determine the corresponding user’s dominant intelligence. The user’s tweet will go through a sentiment analysis and intelligence type classification by applying the SentiStrength method to the Apache Spark framework. This research was conducted on 20 Twitter user accounts with the types of intelligence consisting of musical-rhythmic, visual-spatial, verbal-linguistic, logical-mathematical, bodily-kinesthetics, interpersonal, intrapersonal, naturalistic, and spiritual intelligence. The 72% average accuracy is obtained by calculating the correlation between manual testing and the system using the Spearman’s rank correlation coefficient (α = 0.05 and the Spearman’s value of 0.700).

Humans naturally and intuitively use important and powerful facial expressions to communicate and show their emotions during social interactions. In that interaction, the effort to interpret someone’s emotional condition is important in good communication. A person’s emotional condition is reflected in the form of words, gestures, and especially facial expressions. Although humans can recognize expressions well, research on expression recognition carried out by machines is continuously being carried out in order to be able to carry out expression recognition in human and computer interactions. This research conducted facial expression recognition using the Convolutional Neural Network method. To perform facial expression recognition, the Convolutional Neural Network is trained with expression image data. The training process is carried out using different optimizer values, batch size, and epoch to get the best model. To overcome overfitting, data augmentation was carried out on training data and validation data. The experimental results in this research indicate that the system method built is able to recognize a person’s facial expressions with an accuracy rate of 80%.

Online news portals are one of the main sources of information currently most accessed by the public. With the number of Internet users increasing day by day, online news portal business entrepreneurs must find ways to remain able to compete to attract netizens’ attention to open their news portals. One of the techniques used is to implement clickbait to increase visitor traffic or pageviews. The characteristic of clickbait headlines is that they hide the facts or actual news content in the title section. Clickbait’s goal is to get curious Internet users to click on the link with the headline. This research aims to detect Indonesian-language clickbait headlines automatically by utilizing Indonesian language news headline data obtained from several online news portals with the highest amount of traffic on the Internet. This research uses the Recurrent Neural Network method based on Long Short-Term Memory. The model is built using word embedding to represent text data into vector data. The result of this research is a model with a testing accuracy of 83% and a system that is able to classify headlines whether they are clickbait or not.

The Bcrypt algorithm is a hashing function created from the Blowfish Algorithm by two computer security researchers, Niels Provos and David Mazieres. This hashing function has several advantages, using the original random salt (the salt is the order in which it is added to the password to make it harder to bruteforce). Random salts also prevent lookup table creation. On this basis, the authors try to do a Brute Force experiment on plaintext that has been encrypted by the Bcrypt Algorithm based on 3 characters, namely alphabetic characters, numeric characters and mixed characters to see the security results of the Bcrypt Algorithm. From the results of tests conducted, the alphabetic character with a total of 4 characters can be returned to the original plaintext within 4 days while if the number of 5 characters cannot be found the original plaintext. Then the numeric characters with a total of 7 characters can be found in the original plaintext within 10 hours. Meanwhile, for mixed characters with a total of 7 characters, the original plaintext cannot be found within 5 days. The results of this study indicate that the security performance of the Bcrypt Algorithm is very good in warding off Brute Force attacks for mixed characters while the numeric and alphabetic characters are not good enough.

Gangguan autisme banyak ditemukan pada anak yang berumur 3 tahun ke bawah. Pendiagnosaan gangguan penyakit ini telah dilakukan dengan menggunakan berbagai metode, terutama metode dalam dunia psikologis. Peneliti menggunakan metode Support Vector Machine (SVM) dan metode Naive Bayes untuk menyelesaikan kasus gangguan autisme yang mengalami kesalahan diagnosa. Dalam hasil penelitian ini dilakukan perbandingan metode Support Vector Machine (SVM) dengan metode Naive Bayes. Metode Support Vector Machine (SVM) menghasilkan rata ?¢â?¬â?? rata klasifikasi 93,12%, sedangkan metode Naive Bayes menghasilkan rata ?¢â?¬â?? rata klasifikasi 73,34%.

At this time the development of technology that forces everything must move fast, the efficiency of time becomes a measure of the success of a distribution or production. the continuity of production and the success of production itself is related to distribution. When the distribution fails, production will fail. Indonesian people especially in North Sumatra often consume snacks when free time and snacks are the most consumed food by the Indonesian population until now the production of snacks in Indonesia rose to 50% in 2019, causing the food production to have a concept in timely distribution without experiencing delays in distributing goods that have been produced. One of the expeditions in Indonesia experienced problems in terms of late distribution of goods up to 15-20%, this would have a bad impact on the industry in the future, so to overcome the problems above the application of the ant colony optimization algorithm can help the above problems by placing 4 ants and the ants will spread from point A to point H. In the results of this study, there are 4 ants that move together where the best results are obtained by ant 2, closest to the point H, where the value obtained reaches 0,00015