International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 11 Issue: 04 | Apr 2024
p-ISSN: 2395-0072
www.irjet.net
STOCK MARKET PREDICTION USING HYBRID LSTM-ARIMA AND SENTIMENT ANALYSIS ON NIFTY STOCK INDEX Dr. S. Annie Joice1, U. Mohamed Baashith2, R. Manikandan3, P. Sathasiva Pandi4 1Assistant Professor, Department of CSE, Government College of Engineering, Srirangam, Tamilnadu, India 2,3,4UG student, Department of CSE, Government College of Engineering, Srirangam, Tamilnadu, India
---------------------------------------------------------------------***--------------------------------------------------------------------an era dominated by social media, understanding the impact Abstract - This paper presents a hybrid approach for stock
of public perception on stock prices is paramount. By incorporating sentiment analysis of tweets related to a specific company, the model integrates real-time market sentiment, providing a more holistic perspective. The choice of data sources is pivotal to any prediction model. For the hybrid approach, historical stock market data is sourced, which is then divided into seasonal, residual, and trend components. Concurrently, relevant tweets from Twitter pertaining to the selected company are collected, providing the sentiment analysis algorithm with a rich dataset. This meticulous curation of diverse data sources ensures that the model is not only comprehensive but also capable of adapting to the multifaceted nature of stock market dynamics. This paper is driven by the recognition that accurate stock market predictions empower investors to make informed decisions, mitigating risks and maximizing returns. By embracing a hybrid LSTM & ARIMA model with sentiment analysis, this paper aspires to contribute to this evolving landscape, pushing the boundaries of what is possible in forecasting stock market prices.
market price prediction by integrating Long Short-Term Memory (LSTM), Auto Regressive Integrated Moving Average (ARIMA), and sentiment analysis derived from Twitter data. The proposed hybrid LSTM-ARIMA model involves gathering company-specific data, segmenting it into three components based on trend, seasonal, and residual characteristics, and incorporating tweets for sentiment analysis. The proposed model utilizes LSTM to handle seasonal and residual data, ARIMA for trend data, and sentiment analysis to gauge market sentiment. The hybrid model demonstrates enhanced predictive accuracy compared to standalone LSTM and ARIMA models, underscoring the efficacy of integrating these methods. Tweets obtained from Twitter are utilized to inform decisions regarding stock transactions. Results highlight the significant impact of sentiment analysis on improving overall prediction performance. The hybrid LSTM-ARIMA model outperforms the standalone models and gives an MAE of 0.58, RMSE of 1.16, MAPE of 0.01. The hybrid LSTM-ARIMA model achieves best accuracy for every stock when contrasted with the two standalone models (ARIMA & LSTM). This paper contributes to the advancement of stock market prediction methodologies and recommends the user to buy or sell a particular stock.
This paper investigates into this realm, proposing a hybrid approach [5] that combines LSTM and ARIMA models with sentiment analysis [6]. This integration aims to overcome the limitations of stand-alone models, providing a more robust and reliable prediction framework.
Keywords: Hybrid LSTM-ARIMA Model, Sentiment Analysis, Seasonal, Residual, Trend
The rest of the paper is organized as follows: Section 2 summarizes the related work on stock market price prediction. Section 3 describes the proposed hybrid LSTMARIMA model with sentiment analysis, as well as the training procedure. Section 4 describes the dataset, and its preprocessing for training. Further, the experimental results and performance analysis are also included in this section. In Section 5, the conclusion and future research direction are highlighted.
1. INTRODUCTION The stock market, as a dynamic and complex system, constantly challenges investors and analysts to develop more accurate prediction models. In recent years, machine learning techniques [1] have emerged as powerful tools for forecasting financial markets. Historical stock market prediction methods have predominantly relied on singular models, such as ARIMA or LSTM [2] [3] [4]. This paper aims to bridge the gap between LSTM and ARIMA, two widely used models, by combining them to better understand the complex patterns in financial data. LSTM excels at capturing long-term dependencies, while ARIMA is effective in modeling trend data. This combination leverages the strengths of each model, creating a synergy that promises enhanced forecasting accuracy. In addition to the hybrid modeling approach [5], this paper introduces sentiment analysis as a crucial component [6]. In
© 2024, IRJET
|
Impact Factor value: 8.226
2. RELATED WORK Researchers and practitioners extensively investigated various approaches to develop models capable of forecasting stock prices, addressing the challenging and complex task of stock market price prediction. In their studies, they explored key areas and methods commonly used in this field. Zhao et al. [7] employed time series relational models for stock price prediction, enhancing
|
ISO 9001:2008 Certified Journal
|
Page 845