Search

The emergence of new technologies in the current digital era has led to new challenges and innovative solutions. Researchers are utilizing this revolution for their study, leading to a significant increase in the number of scientific publications globally. Scientific papers typically appear to be complex and difficult to grasp. Understanding the vital phrase in a complex string of terms is quite challenging, but understanding the vital information is crucial for researchers. A crucial factor in determining, whether a publication among the enormous number of publications, is worthy of being accepted is the success of the publications. Peer reviews, which are provided by reviewers who are typically experts in the field, are used in this approach. It is possible to explore scientific peer reviews and foretell the effectiveness of the publications by using Sentiment Analysis (SA, also referred to as opinion mining). Several Machine Learning (ML) algorithms have been used by researchers to analyze scientific peer papers. This literature review seeks to give a comprehensive account of what researchers have done in this field so far and where the current research stands right now. [ABSTRACT FROM AUTHOR]

Many people today struggle with mental health conditions like anxiety and depression. These problems can be resolved with targeted counseling. Depression can cause some people to experience sad thoughts, irregular sleeping habits, a sense of loss of control over their lives, a desire to isolate themselves from friends and family, tension, headaches, and suicidal thoughts. To determine the severity of anxiety and depression, proper screening and monitoring techniques are required. Humans can identify them through patient behaviors, but machines are now able to do it too with the help of Machine Learning (ML) techniques. These techniques help anticipate patients' levels of anxiety and depression and also aid psychologists in treating their patients and in choosing the best course of action based on their level of disease. This paper reviews the studies published between 2010 and 2022. Some keywords such as Machine Learning, anxiety, and depression were used for collection purpose and tools such as Convolutional Neural Network (CNN), Support Vector Machine (SVM), K-Nearest Neighbor, Random Forest, Decision Tree and Naive Bayes were used for analysis, besides deep learning and Machine Learning. SVM and CNN algorithms outperformed other algorithms in terms of accuracy. This study highlights the most recent developments in the field. [ABSTRACT FROM AUTHOR]

The volume of activities and information disseminated daily on the Internet through different social media platforms and channels is increasing geometrically, and we can simply assert that we are approaching an era of connectedness. However, with affordable mobile devices and more Internet penetration, the spread of fake news and misinformation is also emerging as a global social and economic menace. This paper reviews the trends of misinformation in the health sector and the various technological solutions, especially intelligent algorithms, that have been developed to contain this social evil in recent times. The paper adopts the systematic review approach, and the surveyed papers are drawn from ScienceDirect and Scopus databases. The findings reveal that efforts are on to contain the spate of misinformation in the healthcare sector, and that machine learning algorithms are the popular tools currently used by researchers in creating solutions that could deter the spread of fake news and misinformation in the healthcare sector globally. [ABSTRACT FROM AUTHOR]

Fault detection in transmission lines plays an important role in maintaining the dependability and steadiness of power networks. Traditional methods for identifying faults often struggle to handle the diverse nature of real-world fault situations. Machine learning (ML) algorithms offer a data-centered approach that can adjust and learn from datasets, potentially overcoming the limitations of traditional approaches. This paper presents a review of progress in using ML for detecting faults in transmission lines. By drawing insights from lots of studies, explores the paper how ML algorithms have evolved in fault detection, including techniques like networks, recurrent neural networks featuring long short-term memory (LSTM) and convolutional neural networks (CNN). The paper delves into the spectrum of applications where ML is used for fault detection across fault scenarios and operational settings. Additionally, it discuss the obstacles and prospects linked to putting ML-based fault detection systems into practice, such as challenges with data quality, model interpretability and integration with existing grid monitoring systems. Finally, the paper outlines future research paths focused on pushing forward the boundaries of fault detection in power transmission systems through approaches and collaborative endeavors involving academia, industry players and policymakers. In general, this review highlights how ML has the power to revolutionize fault detection methods, enhancing the resilience and dependability of power systems. [ABSTRACT FROM AUTHOR]

This paper explores the implications of marrying Internet of Things (IoT) with artificial intelligence (AI) in the realms of automation and security and what it entails. It presents an innovative application description arising from this technology convergence and its future-generation applications. It is found that AI brings strong cyber security, predictive maintenance and decision-making capabilities to the table due to its advanced algorithms and machine learning (ML) models. IoT has the potential to allow businesses to extract and generate big data, automate processes and monitor in real-time thousands and even millions of devices at once. It is believed that smart city, standalone systems and user experience are quite consumer-friendly. Importantly, the study also assesses facets such as data protection and system interoperability. The results include achieving significant 95.8% accuracy in anomaly detection and 93.4% accuracy in predictive maintenance, showcasing the efficacy of AI models in enhancing system reliability and operational efficiency. Vigorous encryption protocol performance ensures data security, thus addressing issues of privacy. The findings underscore the transformative potential of AI-IoT integration in next-generation automation and security applications. [ABSTRACT FROM AUTHOR]

Deep learning algorithms have enabled generative artificial intelligence (GenAI), which has been increasingly popular in recent years because of its capacity to imitate and produce content that looks human. With its unparalleled capacity to create information, visuals, and even whole stories on its own, GenAI has become a game-changer across a number of industries. OpenAI's modern GenAI framework Sora presents a new method by emphasizing symbolic operations on relational algorithms. Sora stands out as a potentially game-changing technology with broad applications for companies across the globe. Sora's capabilities can not only generate new possibilities but also contribute in the enrichment of knowledge databases in applicable domains. This paper examines Sora's potential in the corporate environment by outlining its salient characteristics, practical uses, and possible economic effects. A few recommendations for possible uses in different business fields such as knowledge management, etc. have been included, based on the predictive abilities. Further, a model has been proposed for refining the image/video/audio output. This study offers scholars, academicians, and business professionals fresh perspectives that could spur more research advancement. [ABSTRACT FROM AUTHOR]

In determining the survival rate of fish eggs, stock evaluation, and monitoring fish population increase, computer usage is replacing the human counting technique. In most cases, counting fish eggs is done in a sequential procedure that requires a lot of time and effort, making it challenging to employ big datasets. Traditional counting methods have been used for many years, but many researchers have pointed out that they have several disadvantages, including requiring more time, exceptional human skills, and excessive human interference. A number of academics are working on developing several strategies to simplify the process of fish counting and assessment. In this paper, we provide a systematic review focusing on several factors which help to automate fish egg counting, using deep learning algorithms and Convolution Neural Network (CNN) from standard databases and research publications. Finally, this paper discusses the most suitable features of fish egg counting, deep learning algorithms, and the future direction of fish egg counting. [ABSTRACT FROM AUTHOR]

The availability of complex financial products has increased considerably over the years, and at the same time, individuals are increasingly taking charge of their own financial security. However, there is very little or no information about whether these individuals have the financial knowledge and skills to navigate this new, increasingly complex economic scenario where stock markets play a vital role in all dimensions like never before. Many saving instruments are now intrinsically linked to financial markets and the retail investor has to understand the whole picture before taking a decision to buy or sell. The paper enumerates the factors influencing stock market participation. Whether these factors include financial literacy or the free flow of information from the capital market or inflation rate, etc., and whether they have a bearing on better investment decisions is something to be explored. The paper attempts to study these factors impacting investment behavior. To better understand literacy and its relation to financial decision making, the relationship between these variables, i.e., literacy and financial planning, and literacy and saving behavior, has been studied using unsupervised Machine Learning (ML) algorithms. The findings could be helpful to investment advisers in relooking their clients’ data and honing their financial consultancy plans. [ABSTRACT FROM AUTHOR]

Virtual currency or cryptocurrency has gained ground in the digital era, prominently due to the universal acceptability of blockchain. This paper applied both machine learning (ML) techniques and finance models to analyze the pattern and volatility of cryptocurrencies. The behavior of various cryptocurrencies was comparatively analyzed on various metrics, and it was found that most of them showed a similar pattern under ML techniques. Further, the closing price of the five selected cryptocurrencies showed that Tether surpassed the other cryptocurrencies which also followed the same trend. Financial model GARCH was applied to understand and forecast the volatility of the selected cryptocurrencies. The study also estimated the spillover effect using R studio software to analyze the short- and long-term impacts of Bitcoin and Ethereum on varied cryptocurrencies. Both ML and finance models showed comparable outcomes. [ABSTRACT FROM AUTHOR]

Logistics management plays a crucial role in supply chain operations, encompassing various tasks such as transportation, inventory management, and order fulfillment. The application of deep learning algorithms has emerged as a promising approach to optimize logistics processes and enhance efficiency. This paper explores the utilization of deep learning algorithms for optimization in logistics management, focusing on transportation routing, warehouse management, and demand forecasting. The study discusses the key deep learning algorithms employed in these areas, including deep neural networks, convolutional neural networks, recurrent neural networks, and reinforcement learning. Furthermore, it presents case studies and empirical evidence that demonstrate the effectiveness of deep learning techniques in improving logistics operations. The paper provides an overview of logistics management and the motivation for employing deep learning algorithms for optimization. [ABSTRACT FROM AUTHOR]

Crop production needs to double by the year 2050 to meet the food requirements of the growing global population. Proper management of fertilization and timely disease control are essential in this endeavor. Machine learning technologies are poised to address these challenges by evaluating various elements, including geographical location, soil acidity, and the presence of essential nutrients such as nitrogen, phosphorus, and potassium. The paper proposes an agro-informatic system that helps farmers in crop prediction, fertilizer recommendation, and disease prediction. The proposed system incorporates critical inputs from farmers, ML algorithms such as RF and thrid-party apps. [ABSTRACT FROM AUTHOR]

The paper develops a novel smart property recommendation system that takes into account crucial parameters of a particular area like pollution level (air and water quality index, and noise), crime rate, crowd cover, average traffic congestion level, proximity to markets, hospitals, schools and transport, happiness index, cleanliness, greenery and site rating, and updates these real-time data continuously through Internet of Things (IoT) devices and artificial intelligence (AI) models to the cloud database. Different machine learning (ML) algorithms recommend properties to the users based on their personalized set of choices, location and budget. The system also generates an alarm to notify the respective authorities if any anomaly is detected. [ABSTRACT FROM AUTHOR]

Affective computing, an interdisciplinary field merging psychology and artificial intelligence, plays a transformative role in adaptive learning. By recognizing and responding to learners' emotions, adaptive systems can personalize learning pathways, enhance engagement and optimize learning outcomes. This paper explores the significance of affective computing in adaptive learning using machine learning, addressing ethical considerations and prospects for creating emotionally intelligent learning environments. [ABSTRACT FROM AUTHOR]

Federated learning (FL) is an emerging tool involving a decentralized machine learning (ML) paradigm. It shows lots of promise encompassing viable solutions to address data privacy and efficiency in the manufacturing industry. Though quite expensive, FL is a key process wherein input settings, data collection and analysis in the manufacturing sector become important. These aspects hinder the advanced ML and data-driven methods, which needs to be addressed through offline training program. FL plays a crucial role in effectively tackling significant issues in digital manufacturing (DM) and its advanced version additive manufacturing (AM). The paper provides a thorough analysis of the application of FL in the context of DM and AM. It also reviews the current state of the manufacturing industry with a focus on the difficulties and potential advantages associated with contemporary production methodologies. [ABSTRACT FROM AUTHOR]

With recent advances in technology, many students and scholars have been tempted to use the Internet as their main educational resource since they can obtain a variety of documents online. These documents can be classified as either formal or informal in writing style, involving different linguistics. The paper presents a method to identify automatically the style of a particular document. First, a dataset of online documents was compiled and preprocessed. Next, features were extracted via a Term Frequency--Inverse Document Frequency vectorizer. Classification models were then built using six classification algorithms. Initially, five Machine Learning algorithms--Random Forest, Decision Tree, Support Vector Machine, Multilayer Perception, and Naïve Bayes--were used. Of these five algorithms, the Random Forest algorithm performed best, obtaining an accuracy of 87.44%, high values for precision and recall, and an F-measure with the lowest error rate. In the second experiment, an Ensemble Learning method was used, whereby a Vote algorithm was used with a combination of the five algorithms. This method obtained an accuracy of 91.96%. The method combines several algorithms. [ABSTRACT FROM AUTHOR]

Monitoring and controlling humidity are very important in different settings. Previous studies have used many Machine Learning (ML) algorithms for prediction. In this paper, we have used Internet of Things (IoT) for collecting and sending the data to the server. For predicting the temperature and humidity, ML is used. DHT 11 (Digital Humidity and Temperature) sensor is used to measure temperature and humidity, while to view this sensor data, Thingspeak IoT server is used. To send this sensor data from time to time, ESP8266 NodeMCU is used, which acts as both microcontroller and wifi module. With this, we can send the data through the Internet and the data can be viewed on the Thingspeak IoT server. Thus, we were able to develop a low-cost system for prediction. [ABSTRACT FROM AUTHOR]

Oral cancer is a significant public health problem worldwide, with early detection and accurate diagnosis being critical factors for successful treatment. Globally, reports of Oral Squamous Cell Carcinoma (OSCC), a prevalent type of cancer, have increased. Histopathological analysis of tissue samples is considered to be the gold standard in the diagnosis of OSCC, but the process is time-consuming and labor-intensive, and requires specialized expertise. Convolution Neural Networks (CNNs), one of the emerging machine learning approaches, have a promising future in automating the identification and diagnosis of cancer using histopathological images. In this paper, two popular deep neural networks, SqueezeNet and GoogleNet, are used to detect OSCC from histopathology images. Each model was trained on different learning rates and their corresponding accuracies were noted. It was found that SqueezeNet gave higher validation accuracy. [ABSTRACT FROM AUTHOR]

This paper proposes a feature optimization and detection method based on the bee scout algorithm and derived support vector machine (DSVM). The DSVM approach reduces network training time by removing unused features. First, the raw EEG data is decomposed using discrete wavelet transform (DWT) into a sequence of frequencies. Before the feature extraction procedure, the spatiotemporal component of the decomposed EEG signal is represented as a twodimensional spectrogram using the shifting. To extract features, four pre-trained SVMs are employed. Dimensional reduction and feature selection are accomplished by bee scout-based EEG channel selection and DSVM approach. The proposed algorithm is tested on MAHNOB dataset. The results suggest that the proposed algorithm is more efficient than existing algorithms. [ABSTRACT FROM AUTHOR]

As cloud computing usage gains popularity, the security vulnerabilities are also increasing. Cloud computing is affected by many types of attacks like account hijacking, data breaches, malicious insiders, human error, etc. This paper primarily focuses on Distributed Denialof-Service (DDoS) attacks, a major security issue for a cloud system. By sending huge amounts of unusual traffic, a DDoS attack corrupts the resources and makes them inaccessible. DDoS attacks are designed to drain resources like main memory, CPU refining, processing space and network frequency, thus making them inaccessible to the customers. It is difficult to identify the difference between the attacker’s requests and legitimate user requests in cloud systems at all stages, especially when the latter comes from a huge number of geographically dispersed workstations. The paper provides a detailed analysis of DDoS attacks and makes a comparative study of various algorithms against the attacks on cloud computing [ABSTRACT FROM AUTHOR]

The largest disruptive component of traffic management system is road transport. Intelligent Transportation Systems (ITS) necessitate exact traffic forecast. However, transport departments continue to struggle to choose an ITS prediction method. A user must be able to appropriately utilize estimation model information. This paper covers contemporary estimation approaches, explains their fundamental concepts and evaluates different traffic estimation approaches for the benefit of decision makers. Each method is classified as Machine Learning (ML), Artificial Intelligence (AI), Deep Learning (DL) or hybrid algorithms. Many surveys are model or datadriven. This study is the first to examine traffic prediction using a variety of algorithmic and methodological approaches to key traffic parameters. Predictive evaluation of dependent variables is done and each algorithm's traffic characteristics are examined in chronological order. The application and effectiveness of each strategy are summarized, and the issues found by analysis are addressed by the review. Hybrid Computational Intelligence (CI)-ML and DL traffic prediction systems are found to be better than previous approaches. Open questions and future possibilities for DL and hybrid traffic prediction systems are also discussed. [ABSTRACT FROM AUTHOR]

A vast range of Neural Network (NN) applications have been reported in recent research. This paper uses a NN-based classification model for dengue disease prediction. Uncertain NN (UNN) is introduced to classify the dengue disease dataset efficiently. Research can be conducted in a wide range of fields to resolve the issue of missing and uncertain data. Realtime datasets contain a lot of missing attributes or values. Real-time data is used because it is approximately measured and thus pays more attention to disease prediction. The main objective is to develop an efficient feature selection-based classification model to predict dengue disease. The research is carried out for feature model creation and relative analysis to improve dengue virus prediction. The work began by gathering dengue virus datasets from the UCI Machine Learning repository, then, a Genetic Algorithm (GA) is used for feature selection; next, the UNN model is used to create a classification model. Accuracy (A), Specificity (Sp) and Sensitivity (Se) are calculated for justifying the result. A simulation of the proposed method is conducted using the WEKA tool. The proposed classification procedure is found to be the most effective model to predict and detect dengue fever. [ABSTRACT FROM AUTHOR]

The paper classifies whether or not a person has brain tumor using transfer learning approach. EfficientNetB0 and VGG16 models were employed because of their accuracy. The results indicate that the accuracy achieved by EfficientNetB0 was more than that of VGG16 model. Due to its high accuracy, transfer learning model can be used for decision-making purpose with respect to clinical scenarios. [ABSTRACT FROM AUTHOR]

Agile Software Development (ASD) is an iterative method of project management and software development that enables teams to provide customer value more quickly with fewer complications. Product Backlog Items (PBI) are a prioritized list to-be-implemented ASD requirements. Prior to beginning sprinting in the ASD, one of the most crucial duties is the classification of PBI. Using Machine Learning and Deep Learning methods, the authors sought to categorize the PBI into Spikes, Foundational stories and User stories. They initially gathered data from numerous software development initiatives and web sources. Each PBI was classified manually as Spikes, Foundational stories, or User stories and preprocessed to remove superfluous text content. Using the TF-IDF and GloVe techniques, the preprocessed PBI were then embedded with words. ANN and LSTM were used to classify the PBIs. The paper models combinations of TF-IDF+ANN, GloVe+ANN and GloVe+LSTM. With an accuracy of 92.4%, the GloVe+LSTM model surpassed other deployed models. [ABSTRACT FROM AUTHOR]

Implementation of smart appliances suitable for smart home and smart cities using modern technologies like Internet of Things (IoT), AI, and Machine Learning is the current trend. Power saving in the design of these new appliances is also a challenge. There are different models of smart fan which can operate at low power having their own merits and demerits. Most of those models work with low power and can control speed automatically with surrounding temperature as a parameter. In this paper, an IoT-enabled innovative smart fan operating on real feel algorithm suitable for smart homes has been developed. Since it is IoT-enabled and works on real feel algorithm, the newly developed fan has some unique superior features different from existing products. [ABSTRACT FROM AUTHOR]

The paper uses three segmentation methods for efficient and accurate detection and classification of breast cancer depending on Region of Interest (ROI). ROI gives maximum information and hence classification is very easy. The approach is compared with existing methods like watershed, region growing method and other existing methods, and gave the best accuracy. The proposed model vastly increased (0.7 to 4.6%) the accuracy of classification and computational time, compared to other models. [ABSTRACT FROM AUTHOR]

The paper aims at predicting whether a person is easily susceptible to heart-related diseases. Most of the research in this area has focused on comparing the commonly known algorithms. A comparison of a majority of algorithms has been done to find the best algorithm. Multiple Machine Learning (ML) classifications were run to determine the outcome. Support Vector Classifier (SVC) and Nu-SVC had highest accuracy in determining heart-related disease. Due to its high accuracy, it was concluded that ML algorithms are best suited for decisionmaking purpose with reference to clinical scenarios. [ABSTRACT FROM AUTHOR]

Image processing in the medical domain has a critical role in diagnosis and decision-making. In automation of medical diagnosis, captured images are processed through various computations to give an earlier and faster medical diagnosis. With the development of new technologies, the process of computing and analysis has widened from a constrained remote processing to worldwide monitoring. Medical diagnosis is rapidly developing due to the advancements in data processing and the demand of faster and accurate decisions. The automation of mammogram is an optimal solution for provisioning faster and early decision in detecting breast cancer. In this paper, fuzzy-based denoising approach is implemented and compared with existing mean and median filters. This approach showed better results. [ABSTRACT FROM AUTHOR]

Bitcoin and Ethereum blockchains are exploding with data. While blockchain is becoming increasingly popular, the sheer volume of data obscures important concerns about safety, and privacy. Data analysis can aid in identifying problems and recommending possible solutions, and is also essential to make blockchain useful in a wide range of industries. The paper reviews the current state of research in four key areas: security, privacy, performance and price prediction, and compiles relevant literature. Finally, it predicts future trends and challenges in this field, which will serve as a roadmap for researchers. Blockchain data analysis will increasingly rely on Machine Learning. [ABSTRACT FROM AUTHOR]

The Internet of Things (IoT), one of the leading cutting-edge innovations, has become an economically attractive field of focus for the scientific community. It requires several system interconnections and device interactions with humans. In order to manage its data exchange and analysis, IoT needs a cloud computing environment; Artificial Intelligence (AI) is needed at the same time via the Internet and cloud-based network of networks. These interconnected IoT systems can interact and share information with each other using their respective identifiers and embedded sensors on each device. We live in the age of big data, and it has become very important to easily and reliably interpret the captured big data. However, while AI is currently playing a greater role in strengthening conventional safety, there are significant challenges to cloud security and IoT computer networking. There are many security concerns that cloud be a danger to the community. In comparison, several of the IoT systems installed on a public network that is wirelessly accessible are now under persistent cyber attack. The paper suggests a hybrid identification paradigm as a response strategy that utilizes Machine Learning (ML) and AI to mitigate and combat IoT cyberattacks both at the host and network levels in cloud computing environments. [ABSTRACT FROM AUTHOR]

Combination of Wireless Power Transfer (WPT) and artificial intelligence (AI) offers huge prospects for several industries such as consumer electronics. This study looks at the recent developments in WPT, particularly concerning AI-improved efficiency, range and reliability. Data based on experiments show a significant improvement in the efficiency of information transfer where AI is used, with rates up to 25% higher than those using traditional ways. Additionally, advanced maintenance forecasting algorithms improve system availability by cutting down on 30% of downtimes in real-life applications of AI technology. The study also looks at the scope for further research, including emergence of distributed AI for autonomous WPT network, and quantum computing for improving algorithms. The study concludes that the convergence of WPT and AI could enhance the efficiency and robustness of power transfer systems, and hence it is essential to continue the investigation in this interdisciplinary field to advance existing technology and discover novel possibilities. [ABSTRACT FROM AUTHOR]

This study investigates how household inflation expectations relate to realized inflation, and uses biased expectations hypothesis (BEH) to understand the association between inflation expectations and realized inflation in the food sector. Household survey data, wholesale price index (WPI), and consumer price index (CPI)-based data on inflation are used in the study and the data was sourced from the Reserve Bank of India (RBI) and Ministry of Statistics and Program Implementation, Government of India. The obtained results were cross-validated using auto-SARIMAX algorithm in a machine learning (ML) framework. The study finds that expected and realized inflation have significant causal relationships; this further supports BEH and reveals how household inflation expectations impact actual market inflation, and vice versa. Hence, these are broadly complementary. However, changes may occur in these aspects, and households and markets must prepare for these changes and plan their future accordingly. [ABSTRACT FROM AUTHOR]

Vehicle detection and classification is an active area of research in vehicle surveillance system and has numerous applications in intelligent transportation system. In vehicle surveillance system, identifying the make and model of a vehicle is crucial for traffic monitoring. Due to intraclass diversity, viewpoint variation, and variable illumination conditions, identifying vehicle make and model is a difficult process. Earlier researchers have used different datasets for the detection and classification of vehicle as per their make and model. However, a majority of these studies have used datasets of clear high-quality images. The present study uses dataset with vehicle images taken under uncontrolled environmental conditions such as images with different illumination, shadowing, reflection from vehicle surface, etc. and employs transfer learning approach to develop an efficient algorithm for computer vision based on traffic surveillance system that can detect and classify vehicles. The study further compares and analyzes the accuracy and precision of the two transfer learning techniques. The study deploys machine learning (ML) classifiers for classifying the vehicles as per their models. [ABSTRACT FROM AUTHOR]

The case discusses the failure of International Business Machines’ (IBM) artificial intelligence (AI) software Watson for Oncology (Watson). In 2012, the American multinational technology company partnered with New York-based cancer treatment center and research organization Memorial Sloan Kettering Cancer Center (MSK) to develop Watson that could provide medical professionals with improved access to up-to-date and comprehensive cancer data and practices. Despite Watson’s Natural Language Processing (NLP) ability that enabled it to read and gain insights from unstructured data it was not able to interpret data as human doctors could. A 2017 investigation carried out by news website STAT revealed how IBM’s AI software could not live up to the hype created around it by the company. To add to IBM’s troubles, a 2017 audit carried out by the University of Texas showed that MD Anderson was using old data to train its OEA. The same year, i.e. in 2017, the cancer center closed down its project with IBM after spending $62 mn. With data quality and domain expertise being some of the major reasons for the failure of AI projects, is it worth it for companies such as IBM to invest in AI and Machine Learning (ML) tools in healthcare like Watson for Oncology without proper data preparation? This study looks at the roadblocks standing in the way of IBM finding success in using AI and ML in a bid to drive clinical research and drug discovery. It also examines how technology companies can make AI and ML a truly transformative force in healthcare. [ABSTRACT FROM AUTHOR]

Application of artificial intelligence (AI) and machine learning (ML) has been gaining wide currency in various domains, ranging from streaming video, online shopping recommendations, image detection and generation, autonomous car driving, and also finance. The Indian AI fintech market is predicted to grow by 1.5-3 times by 2025, according to some estimates. AI is the ability of machines to simulate human behaviors, including decision-making. ML is the ability of machines to automatically learn without being explicitly programmed. ML algorithms learn with the data provided to them and become better as more and more data is provided, and the accuracy of the model for prediction or classification improves drastically if the right set of data is provided. This study gives an overview of how ML can be used in investment management along with use cases. [ABSTRACT FROM AUTHOR]

Telemarketing is an integral part of persuasive technology-driven commerce tools that have progressed by virtue of advancements in the digital technology. Currently, telemarketing is adopted for modern commerce settings. The prominence of telemarketing promotion cannot be avoided as it has a positive influence on the overall economy. The current study uses various data modeling techniques to predict the achievement of telemarketing promotion. Data was collected for the study from a Portuguese retail bank from 2008 to 2013. The data consists of 21 features, associated with bank client, product and social-economic attributes that were examined. The study also uses some advanced algorithms that have not been used previously, to forecast the reaction to the telemarketing campaign. The study scrutinizes: 1) which algorithms perform better; and 2) which parameters are superior indicators of the models' performance. The findings reveal that the best Machine Learning model for the telemarketing campaign success prediction is Decision Tree with F1-score of 0.5936 and L-measure of 0.5634. Bank managers should use the Decision Tree model to gain definite insights on the behavior of prospective bank customers. [ABSTRACT FROM AUTHOR]

Petroleum products fuel the economic engine of a country. It is vital that accurate demand forecasting is done for these products. Various forecasting methods, from simple methods like moving average to complex fuzzy logic, have been used to forecast the demand for petroleum products with varying degrees of accuracy. This study compares the forecasting accuracy of two Machine Learning forecasting models, namely, Seasonal Auto Regressive Integrated Moving Average (SARIMA) and Neural Network, to forecast the consumption of gasoline (MS) and diesel (HSD) in India and find out which model is able to better predict the demand. To compare the forecast accuracy of models, Mean Absolute Percentage Error (MAPE) is used. The model with the lowest MAPE is considered the better forecasting model. The study concludes that SARIMA and Neural Network are able to predict the consumption of gasoline (MS) with almost equal accuracy, while SARIMA is able to predict the consumption of diesel (HSD) significantly better than Neural Network. [ABSTRACT FROM AUTHOR]

In response to the broken agriculture supply chain caused by a lack of storage facilities, lack of transportation, and manifold intermediaries and lack of transparency and traceability, Thirukumaran Nagarajan launched Ninjacart, a leading agritech firm in India, in 2015, along with Sharath Babu Loganathan, Ashutosh Vikram, Kartheeswaran K K and Vasudevan Chinnathambi. After exploring the issues in the agriculture sector, Ninjacart shifted its operations from B2C to B2B, to leverage the profitable business opportunity. Ninjacart used advanced level supply chain algorithms that it developed in-house, predictive analytics, Big Data, IoT technology, and mobile applications to help farmers overcome the various issues they were grappling with. As of early 2020, Ninjacart became one of the largest fresh produce supply chain companies in India. With the Covid-19 pandemic leaving millions of farmers across India in the lurch, Ninjacart launched many initiatives. As part of its future plans, Ninjacart was banking on Machine Learning to improve operations. However, it faced several challenges. Having established itself as a prominent player, Ninjacart's next challenge was scaling up its operations while improving efficiency and profitability at the same time. [ABSTRACT FROM AUTHOR]