ItemPhD

A Comprehensive Model for Forecasting the Nifty50 Index Using MAchine and Deep Learning Methodologgy with Reference to National Stock Exchange

Title
A Comprehensive Model for Forecasting the Nifty50 Index Using MAchine and Deep Learning Methodologgy with Reference to National Stock Exchange
Creator
Chinthakunta, Manjunath
Contributor
M, Balamurugan and Ghosh, Bikramaditya
Description
The volatility and uncertainty make stock and stock price index predictions challenging. Many financial professionals and academics are interested in stock price/index prediction studies. This study presents computational ML and DL intelligence techniques for estimating the NIFTY50 index closing value on the Indian NSE using Fundamental Analysis and Technical Analysis. To forecast the NIFTY50 index, we first employed Fundamental Analysis and max voting, bagging, boosting, and stacking ensemble learning techniques. An embedded feature selection algorithm is utilized to determine the model's best fundamental indicators, and a grid search is performed to tweak hyperparameters for each base regressor. Our results demonstrate that the bagging and stacking regressor model 2 beat all other models, with the lowest RMSE of 0.0084 and 0.0085, respectively, indicating an improved fit of ensemble regressors. Subsequently, TA research was done to exhibit the influence of deep learning on the NIFTY50. This method employs a data augmentation mechanism and three GRU model variations. It is examined using two datasets, TA1 and TA2, which include technical indicators from the NIFTY50 index. The GRU model enhanced the NIFTY50 index prediction using the TA1 technical indicator dataset. Finally, the study examines a hybrid model to estimate equity market trends, combining PCA with ML methods such as ANN, SVM, NB, and RF. The proposed approach uses the trend deterministic data preparation layer to convert the continuous data to a discrete form denoted by +1 or -1. The empirical findings of this hybrid model demonstrate that the RF model with the first three principal components obtains precision of 0.9969, F1-score 0.9968 and AUC score of 1. Overall, the suggested research design outperforms baseline models in our experiments and shows promising results using fundamental and technical analysis indicators. Thus, this study provides an ideal tool for stock market prediction and financial decision-makers.
Source
Author's Submission
Date
2024-01-01
Publisher
Christ(Deemed to be University)
Subject
Computer Science and Engineering
Rights
Open Access
Relation
61000387
Format
PDF
Language
English
Type
PhD

Collection

PHD

Citation

Chinthakunta, Manjunath, “A Comprehensive Model for Forecasting the Nifty50 Index Using MAchine and Deep Learning Methodologgy with Reference to National Stock Exchange,” CHRIST (Deemed To Be University) Institutional Repository, accessed February 23, 2025, https://archives.christuniversity.in/items/show/12433.

Output Formats