ArthaYukti: An Integrated Deep Learning Framework for Financial Sentiment Analysis and Stock Price Forecasting

Abstract

This paper presents ArthaYukti, a unified deep learning framework that integrates financial sentiment analysis with stock price forecasting to generate actionable market intelligence. The system employs FinBERT for domain-specific sentiment classificationandLongShort-Term Memory(LSTM) networks for time-series forecasting. A dynamic weighting mechanism adaptively fuses sentiment- driven signals with technical indicators, thereby improving predictive accuracy over standalone models. Experimental results demonstrate notable reductions in forecasting error, improved sentiment detection, and efficient real-time processing. The proposed platform facilitates enhanced decision making for investors and financial institutions.

Introduction

Financial markets are complex and influenced by both quantitative factors and qualitative information such as news and public sentiment, which traditional technical and fundamental analysis often fail to capture together. Recent advances in deep learning and NLP—particularly LSTM networks for time-series modeling and transformer-based models like BERT and FinBERT for financial text—have improved market analysis, but most existing systems treat sentiment analysis and price forecasting as separate tasks, limiting their effectiveness.

ArthaYukti addresses this gap through a hybrid deep-learning framework that jointly integrates financial news sentiment and historical stock data. The system combines a real-time data pipeline, a FinBERT-based sentiment analysis module, an LSTM-driven stock price forecasting model, and a dynamic weighting mechanism that adaptively fuses sentiment and technical indicators based on market conditions. This allows the model to shift emphasis toward sentiment during volatile periods and toward technical signals during stable markets.

The platform uses a modular architecture with real-time market and news data acquisition, efficient backend processing, and an interactive frontend for analytics. Experiments on five years of NIFTY50 data show strong performance: FinBERT achieved about 90% accuracy in sentiment classification, the LSTM model produced low forecasting errors, and the hybrid approach improved prediction accuracy by over 6% compared to LSTM-only models. System tests also demonstrated low latency, scalability, and positive user feedback.

Conclusion

ArthaYukti introduces a novel hybrid approach integrating financial sentiment analysis with LSTM- based stock forecasting. The system’s adaptive weighting algorithm, real-time ETL pipeline, and scalable architecture position it as a valuable tool for modern financial analytics. The results confirm that combiningsentimentandtechnicalfactorsyieldsmore accurate and reliable market predictions.

References

[1] S.HochreiterandJ.Schmidhuber,“LongShort-Term Memory,” Neural Computation, vol. 9, no. 8, pp. 1735–1780,1997. [2] J. Devlin, M.-W. Chang, K. Lee, and K. Toutanova, “BERT:Pre-trainingofDeepBidirectionalTransformersfor Language Understanding,” Proc. NAACL-HLT, 2019. [3] Y. Yang, M. H. Chen, and A. Hagen, “FinBERT: A PretrainedLanguageModelforFinancialCommunications,” arXiv:2006.08097, 2020. [4] M. Schumaker and H. Chen, “Textual Analysis of Stock MarketPredictionUsingFinancialNewsArticles,”Decision Support Systems, vol. 50, no. 2, pp. 164–174, 2011. [5] X.Ding,Y.Zhang,T.Liu,andJ.Duan,“DeepLearning forEvent-DrivenStockPrediction,”Proc.IJCAI,pp.2327– 2333, 2015. [6] E. F. Fama, “Efficient Capital Markets: A Review of TheoryandEmpiricalWork,”TheJournalofFinance,vol. 25, no. 2, pp. 383–417, 1970. [7] Z. Zhang, S. Wang, and B. Liu, “Deep Learning-Based StockPricePrediction:ASurvey,” IEEEAccess,vol.8,pp. 216139–216152, 2020. [8] A.Graves,“SupervisedSequenceLabelling,”Studiesin Computational Intelligence, Springer, 2012. [9] X.Liangetal.,“IntegratingSentimentwithTechnical IndicatorsforStockPrediction,” IEEEAccess,vol.8,pp. 175307–175322, 2020. [10] Y.Qinetal.,“ADual-StageAttention-BasedRecurrent Neural Network for Time Series Prediction,” Proc. IJCAI, 2017. [11] H. Liu, L. Wang, and J. Zhang, “An Improved Deep LearningModelforStockIndexPrediction,”IEEEAccess, vol. 7, pp. 73110–73118, 2019. [12] H.G.KimandC.H.Kim,“FinancialNewsAnalysis for Stock Market Prediction Using Machine Learning,” Expert Systems with Applications, vol. 130, pp. 157–168, 2019. [13] S.Liao,“DeepLearningforStockPredictionUsing Fundamental Analysis,” International Journal of Data Science, vol. 5, no. 2, pp. 113–129, 2020. [14] A.Vaswanietal.,“AttentionisAllYouNeed,” Proc. NeurIPS, 2017. [15] K.Heetal.,“DeepResidualLearningforImage Recognition,” Proc. CVPR, 2016. [16] Y.LeCun,Y.Bengio,andG.Hinton,“DeepLearning,”Nature,vol.521,pp.436–444, 2015. [17] R.P.Schumaker,Y.Zhang,C.-N.Huang,andH.Chen, “Evaluating Sentiment in Financial News Articles,” Proc. HICSS, 2012. [18] P.C.Tetlock,“GivingContenttoInvestorSentiment: The Role of Media in the Stock Market,” The Journal of Finance, vol. 62, no. 3, pp. 1139–1168, 2007. [19] L.Bing,“SentimentAnalysisandOpinionMining,” Synthesis Lectures on Human Language Technologies, Morgan & Claypool, 2012. [20] J. Patel, S. Shah, P. Thakkar, and K. Kotecha, “Predicting Stock Market Index Using LSTM,” ExpertSystemswithApplications,vol.42,pp.10680–10690,2015. [21] J.Huetal.,“StockMovementPredictionUsingHybrid Attention-Based LSTM,” IEEE Access, vol. 7, pp. 178128– 178139, 2019. [22] T.FischerandC.Krauss,“DeepLearningwithLong Short-Term Memory Networks for Financial Market Predictions,”EuropeanJournalofOperationalResearch, vol. 270, no. 2, pp. 654–669, 2018. [23] M.Z.Aslam,S.Kang,andK.J.Lee,“Sentiment-Based Stock Price Prediction Using Neural Networks,” International Journal of Forecasting, vol. 37, pp. 1520– 1532, 2021. [24] Z. Li, X. Wang, and D. Zhou, “Multi-Modal Stock ForecastingwithSentimentandTechnicalIndicators,”IEEE Transactions on Knowledge and Data Engineering, 2022. [25] J.Bollen,H.Mao,andX.Zeng,“TwitterMoodPredicts the Stock Market,” Journal of Computational Science, vol.2, no. 1, pp. 1–8, 2011.

Copyright

Copyright © 2025 Ganesh Bhutkar, Aditya Deore, Shreyas Done, Pornima Dokhale, Esha Manhas. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.