Abhishek Dutta

Convolution Neural Network is a deep learning method that is broadly used for image classification, image recognition, object detection, etc. •In this project, we will see how CNN’s can be used for image classification. For this, the model takes an image as input, processes it, and classifies it under a certain category. An image is a collection of pixels, with features that specify the height, width, and the dimensions of the image.

(PDF) Handwritten Digits Recognition using Convolution Neural Networks. Available from: https://www.researchgate.net/publication/344417265_Handwritten_Digits_Recognition_using_Convolution_Neural_Networks [accessed Oct 15 2020].

The purview of stock price analysis largely depends on the ability to identify the movement of the stock prices and predict the hidden patterns and trends which the market follows. The sole idea is to gain profit from the investments that we make, therefore the more sure we will be with our predictions, safer will be the outlay. Predictions based on stock prices has been a constant field of research work in the past, however, obtaining the desired level of precision is still an engaging challenge. In this script, we are proposing a combined effort of using efficient machine learning techniques coupled with a deep learning technique (like LSTM) to use them to predict the stock prices with a high level of accuracy. We are considering the daily index values of three different companies namely HDFC bank, Tata Consultancy Services, Cipla which are from different segments of the market-finance, IT and medical science. We are using their daily data of previous 6 years (2013-18) to prepare a training model and implement the results on the test data set to predict the closing values of these National Stock Exchange (NSE) listed companies from January 1 to December 31, 2019. For prediction of the patterns of the price movements we are using efficient classification techniques, and for the actual closing values we are using various techniques of regression. Methods which have been implemented involve logistic regression, SVM (Support Vector Machines), ANN (Artificial Neural Network), Random Forest, ensemble learning techniques (Bagging, Boosting). We also use the deep learning technique of Long Short-Term Memory (LSTM) for the prediction of the closing prices of the stocks and then superimpose the accuracy measures by comparing the LSTM results with the other machine learning models.

This presentation is on stock price prediction using machine learning and deep learning models. Eight machine learning and four deep learning -based models using LSTM have been proposed and these models are evaluated on their execution time and performance accuracies.

The purview of stock price analysis largely depends on the ability to identify the movement of the stock prices and predict the hidden patterns and trends which the market follows. The sole idea is to gain profit from the investments that we make, therefore the more sure we will be with our predictions, safer will be the outlay. Predictions based on stock prices has been a constant field of research work in the past, however, obtaining the desired level of precision is still an engaging challenge. In this script, we are proposing a combined effort of using efficient machine learning techniques coupled with a deep learning technique ( like LSTM) to use them to predict the stock prices with a high level of accuracy. We are considering the daily index values of three different companies namely HDFC bank, Tata Consultancy Services, Cipla which are from different segments of the market – finance, IT and medical science. We are using their daily data of previous 6 years (2013-18) to prepare a training model and implement the results on the test data set to predict the closing values of these National Stock Exchange (NSE) listed companies from January 1 to December 31, 2019. For prediction of the patterns of the price movements we are using efficient classification techniques, and for the actual closing values we are using various techniques of regression.  Methods which have been implemented involve logistic regression, SVM (Support Vector Machines), ANN (Artificial Neural Network), Random Forest, ensemble learning techniques (Bagging, Boosting).  We also use the deep learning technique of Long Short-Term Memory (LSTM) for the prediction of the closing prices of the stocks and then superimpose the accuracy measures by comparing the LSTM results with the other machine learning models.

Stock market indexes predictions have always been under the radar of stalwarts belonging from the domains of econometrics, statistics, and mathematics. This has been a fascinating challenge to deal with since a major portion of the research community who promotes the idea of the efficient market hypothesis (EMH) believes that no predictive model can accurately predict the fluctuations of the ever-changing market, while recent works in this field using more advanced techniques like statistical modelling and machine learning can be used to demonstrate the gesticulations of a time series data with exceptional levels of accuracy. The stock market of a given country can be divided into its constituent sectors which represent the entire behaviour of a particular domain instead of the performance of individual companies.
In this paper, it has been proposed how machine learning and deep neural network technique like Long Short Term Memory (LSTM) can be used to obtain fantastic results for prediction of stock values of any sector over a period of one week using prior data of five days from the previous week. Here, the IT sector of India has been taken into account to analyse its characteristic features about its ascend and descend according to the trend of the market. Regression techniques have been used to predict the probable indexes of the closing values and classification methods for identifying their pattern of movement. At first, a detailed machine learning approach has been adopted by using all adept methods like ensemble techniques like bagging and boosting, random forest, multivariate regression, decision tree, support vector machines, MARS, logistic regression and artificial neural networks. Both regression and classification have been performed on the IT sector data.