Advances in Science and Technology Vol. 124

Abstract: A Stock refers to a share of ownership in a particular company. An investor owns 1 percent of the company if they purchase one part of its ownership divided into 100 parts, each equal to one share. Stock exchanges are run by an automated matching system driven by order demand. Stock prices are defined as “at any particular time, how many buyers and sellers available for the same stock in the market. If the number of buyers is more than sellers, then stock price becomes high” and vice-versa. The stock market depends on many factors like open price, close price, high and low price. Many researchers have tried to predict the stock prices using various ML (machine learning) techniques such as ARIMA model, linear regression, RNN, etc. Because of the uncertainty in stock market, simple models cannot yield any genuine results. The limitations with models like ARIMA, TSLM have been traced in this paper. This paper builds a web application using a library named Streamlit and integrate the stock prediction model. The main objective of this paper is to build a LSTM (Long-Short Term Memory) based RNN (Recurrent Neural Network) model using opening prices in order to get one of the most accurate stock rates.

Abstract: Today’s markets are rather matured and arbitrage opportunities remain for a very short time. The main objective of the paper is to devise a stock market ontology-based novel trading strategy employing machine learning to obtain maximum stock return with the highest stock ratio. The paper aims to create a dynamic portfolio to obtain high returns. In this work, the impact of the applied machine learning techniques on the Chinese market was studied. The problem of investing a particular total amount in a large universe of stocks is considered. The Chinese stocks traded on Shanghai Stock Exchange and Shenzhen Stock Exchange are chosen to be the entire universe. The inputs that are considered are fundamental data and company-specific technical indicators unlike the macroscopic factors considered in the existing systems. In the stock market document repository, ontological constructs with Word Sense Disambiguation (WSD) algorithm improve the conceptual relationships and reduce the ambiguities in Ontological construction. The machine learning techniques Kernel Regression and Recurrent Neural Networks are used to start the analysis. The predicted values of stock prices from the Artificial Neural Network provided quite accurate results with an accuracy level of 97.55%. In this study, the number of nodes will be selected based on Variance-Bias plots by tracking the error on the in-sample data set and the validation data set.

Abstract: This research proposes a railway crack detection system. This research describes a classification system that can classify any crack in the railway tracks by using deep learning with convolutional neural networks (CNN). In the railway network of the Indian railways, accidents are one of the major concerns due to the unidentified cracks that are available on the rail tracks. The majority of accidents occur due to railway track cracks, resulting in the loss of precious lives and economic loss. So, it has become necessary to monitor the health condition of the track regularly by using a train track crack classification system. This project prevents the train derailment by classifying cracks in the railway tracks using image processing technologies. To identify the train track crack classification system that uses deep learning with Convolutional Neural Network architecture of different layers along with certain image pre-processing methods has been very successful in the classification of railway track crack has occurred or not. In convolutional neural network, there are a lot of layers available where training of the images are done which are available in the dataset and these layers are made up of lots of neurons. So its have been found that these convolutional neural networks are considered to be able to record the colours and textures of lesions related to corresponding railway track cracks, which is similar to human decision-making.

Abstract: In a recent survey, its observed that there is arise in crime rate in India, and due to this, many peoplefeel unsafe in this country. So to reduce it, predicting thecrime before it happens is very important. IndianGovernment uses a software called CCIS (Crime CriminalInformation System), this software is only used to storeinformation, but with the stored information, it doesn’t doanything else, the existing systems predict crime only on aday basis, but it doesn’t predict exactly at what hour is itgoing to occur, to predict it on a hour basis, accuracy isimportant, so to analyze and predict crime, the accuraciesof machine learning algorithms such as KNN, Decision treeand Random forest are compared in order to use the bestfor analysis and prediction.

Abstract: The worldwide society was devastated by the 2019 coronavirus illness (COVID19) epidemic in Wuhan, China, which overloaded advanced medical systems around the world. The World Health Organization (WHO) is constantly monitoring and responding to the pandemic. The current rapid and exponential development in patient numbers necessitates the use of AI technology to forecast possible outcomes of infected individuals in order to provide suitable therapy. The goal is to find the machine learning-based solution that best fits the Covid19 vaccination predictions with the highest accuracy. Variable identification, univariate analysis, bivariate and multivariate analysis, missing value handling and data validation analysis, data cleaning / preparation, and data validation analysis are all accomplished using supervised machine learning technology (SMLT). Various types of data, such as visualisation, are gathered. For the entire given dataset. Proposal of a machine learning-based method for accurately predicting the suitability of Covid19 vaccine prediction.

Abstract: Employee attrition rate in Tech industry has become dreadful day by day in all over the world. Meanwhile It has been noticed that churn (attrition) rate in IT industries is growing rapidly than expected especially during pandemic times. This is taken as a foremost issue by each tech industry, to analyze and adapt to the change. The main snag is that, the expenditure of recruiting on a new employee is foremost ineffective than retaining a company trained professional employee. Also retaining an employee will assure certain credibility and work culture of the company than the new employee. Also, the latter will be given access to training modules and code of conduct of the company with lots of Information Overload on a short span of time. It is essential to mention, not every organization has comprehensive training programs for their employees, especially the start-up tech firms, which focuses heavily on skilled workers with experience beforehand. This anonymity causes HR departments to scrutinize and tweak their actions according to current trend in the market. The major goal of this study is to make predictions whether the skillful employee will quit or continue further and predict the reason for quit using supervised classification and machine learning algorithms. Acquainting the human resource team to help them with the required analytics to make decisions based on machine learning.

Abstract: Recommender Systems (RS) systems help users to select items and recommend useful items to target customers who are interested in them, such as movies, music, books, and jokes. Traditional recommendation algorithms are primarily concerned with improving performance accuracy; as a result, these algorithms prefer to promote only popular products. Variability is also an important inaccurate number of personalized recommendations that suggest unfamiliar or different things. Multi objective development strategies, which magnify these contradictory measures simultaneously, are used to measure accuracy and variability. Existing algorithms have an important feature because they are not flexible enough to control competing targets. We suggest creating a recommendation system based on shared filtering. Instead of finding out the preferences and preferences of users openly, we can find out by publicly analyzing historical and real-time data. This is done through a process called matrix factorization. Matrix factorization algorithms work by decomposing the interactive matrix of the user object into a product of two rectangular matrices with a minimum size. This type of recommendation has the added advantage of finding invisible and unmeasured relationships that are not possible with standard content-based filters.

Abstract: One of the major contributor leading to cause of unintentional injuries after motor vehicle crashes and poisoning, is Falls. The existing Fall Prediction Algorithms are used to predict falls in older or disabled people by analyzing their fall history, capturing their movements through visual sensors (cameras, thermal imaging etc.) in a fixed environment, using inertial sensors to identify the patterns of movements. These algorithms are monologues for each person as they learn from their history and predict falls specific only to that person. The algorithm proposed in this paper aims to predict falls using kinematic data such as accelerometer, magnetometer, and gyroscopic values, for any user. This work involves developing an algorithm capable of predicting falls and to achieve this, we use Long Short-Term Memory (LSTM). The benefit of this algorithm is to prevent trauma to the body or at least reduce the impact of fall and the fatality caused by it. In the future, this algorithm can be used to design a device to predict falls in real-time to scenario be used by everyone irrespective of gender, age, and health.

Abstract: Coronavirus (COVID-19) has grown to be one of the most dangerous and acute illnesses in recent years, and it has now spread across the globe. In order to prevent COVID-19, early detection of the Coronavirus is necessary. Using a convolutional neural network (CNN) and long short-term memory (LSTM), we have suggested a model for automatically diagnosing COVID-19 from X-ray images. In this model, CNN is used to extract deep features, while LSTM is utilized to identify those features. The proposed method can aid in the diagnosis and treatment of patients with COVID-19. As a final step, this technology will be able to accurately detect the severity of the disease in the lungs and provide it with an automated diagnostic. This model will be hosted on the website so that hospital visits may be minimized and diagnosis can be delivered at home, if necessary, thereby giving a solution for COVID-19 containment.

Abstract: Using Machine Learning and / or Deep Learning for early detection of diseases can help save people’s lives. AI has already been making progress in healthcare as there are newer and improved software to maintain patient records, produce better imaging for error free diagnosis and treatment. One drawback working with real-life datasets is that they are predominantly imbalanced in nature. Most ML and DL algorithms are defined keeping in mind that the dataset is equally distributed. Working on such imbalanced datasets cause the models to end up having high type-1 and type-2 error which is not ideal in the medical field as it can misdiagnose and be fatal. Handling class imbalance thus becomes a necessity lest the ML/DL model fails to learn and starts memorizing the features and noises belonging to the majority class. PIMA Dataset is one such dataset with imbalances in classes as it contains 500 instances of one type and 268 instances of another type. Similarly, the Wisconsin Breast Cancer (Original) Dataset is also a dataset containing imbalanced data related to breast cancer with a total of 699 instances where 458 instances are of one class (Benign tumor images) while 241 instances belong to the other class (Malignant tumor images). Prediction/detection of onset of diabetes or breast cancer with these datasets would be grossly erroneous and hence the need for handling class imbalance increases. We aim at handling the class imbalance problem in this study using various techniques available like weighted class approach, SMOTE (and its variants) with a simple Artificial Neural Network model as the classifier.