Home / Current Issue / Paper 1717317
Deepstock : AI Powered Stock Analysis
Subject area: Science,Engineering and Technology · Area of research: Machine Learning in Finance / Stock Prediction
DOI: https://doi.org/10.64388/IREV9I11-1717317
Abstract
The pervasive challenge in modern quantitative finance lies in the failure of traditional models to integrate crucial, non-numerical market factors, leading to poor predictive stability during event-driven volatility. This manuscript details the DeepStock platform, an operative architecture engineered to systematically harmonize structured financial time-series data with highly descriptive, unstructured textual narratives. The system leverages the superior contextual reasoning of the Google Gemini Large Language Model (LLM) to generate robust, quantifiable financial sentiment scores. This sentiment feature vector is fused with traditional technical indicators via an Early Fusion strategy to construct a comprehensive predictive feature set. We hypothesize that models incorporating these hybrid features will yield superior risk-adjusted returns, as validated by the Sharpe Ratio, relative to purely time-series-dependent strategies. The deployment utilizes a Flask/Python framework to provide a free, real-time AI tool focused on analysis and prediction within the NSE and BSE stock markets.
References
[1] DeepStock - Free AI Stock Analysis & Prediction Tool | Real-Time ..., accessed November 4, 2025, https://deepstock.in/
[2] Neural based event-driven stock rally prediction using SEC filings and Twitter data, accessed November 4, 2025, https://web.stanford.edu/class/archive/cs/cs224n/cs224n.1204/reports/custom/15787311.pdf
[3] From Deep Learning to LLMs: A survey of AI in Quantitative Investment - arXiv, accessed November 4, 2025, https://arxiv.org/html/2503.21422v1
[4] Combining Financial Data and News Articles for Stock Price Movement Prediction Using Large Language Models - arXiv, accessed November 4, 2025, https://arxiv.org/html/2411.01368v1
[5] On Assessing the Performance of LLMs for Target-Level Sentiment Analysis in Financial News Headlines - MDPI, accessed November 4, 2025, https://www.mdpi.com/1999-4893/18/1/46
[6] [2412.19245] Sentiment trading with large language models - arXiv, accessed November 4, 2025, https://arxiv.org/abs/2412.19245
[7] Building Scalable AI Solutions with Python and Flask - Shiv Technolabs, accessed November 4, 2025, https://shivlab.com/blog/building-scalable-ai-solutions-python-flask/
[8] Transforming a Financial Model into a Web App with AI (Updated Sept 2025), accessed November 4, 2025, https://www.adventuresincre.com/financial-model-web-app/
[9] [2510.15929] Comparing LLMs for Sentiment Analysis in Financial Market News - arXiv, accessed November 4, 2025, https://www.arxiv.org/abs/2510.15929
[10] CatMemo at the FinLLM Challenge Task: Fine-Tuning Large Language Models using Data Fusion in Financial Applications - ACL Anthology, accessed November 4, 2025, https://aclanthology.org/2024.finnlp-2.19.pdf
[11] Can LLM-based Financial Investing Strategies Outperform the Market in Long Run? - arXiv, accessed November 4, 2025, https://arxiv.org/html/2505.07078v3
[12] The architecture of the Multi-modal fusion module. The architecture of... | Download Scientific Diagram - ResearchGate, accessed November 4, 2025, https://www.researchgate.net/figure/The-architecture-of-the-Multi-modal-fusion-module-The-architecture-of-the-Multi-modal_fig1_352279961
[13] Free Download Multi-Modal Data Fusion Architecture Plan - Meegle, accessed November 4, 2025, https://www.meegle.com/en_us/advanced-templates/precision_medicine/multi_modal_data_fusion_architecture_plan
[14] Prompt Engineering for Finance 101 | Deloitte US, accessed November 4, 2025, https://www.deloitte.com/us/en/services/consulting/articles/prompt-engineering-for-finance.html
[15] The Advantages of Using Streamlit, Flask, and Django for Building Web Apps - Medium, accessed November 4, 2025, https://medium.com/@akratech/the-advantages-of-using-streamlit-flask-and-django-for-building-web-apps-67f0907861e4
[16] What is the query limit on Yahoo's Finance API? - Stack Overflow, accessed November 4, 2025, https://stackoverflow.com/questions/9346582/what-is-the-query-limit-on-yahoos-finance-api
[17] Why yfinance Keeps Getting Blocked, and What to Use Instead | by Trading Dude - Medium, accessed November 4, 2025, https://medium.com/@trading.dude/why-yfinance-keeps-getting-blocked-and-what-to-use-instead-92d84bb2cc01
[18] Newspaper3k: Article scraping & curation — newspaper 0.0.2 documentation, accessed November 4, 2025, https://newspaper.readthedocs.io/
[19] How To Scrape News Articles with Newspaper3k (Python) - ScraperAPI, accessed November 4, 2025, https://www.scraperapi.com/blog/python-newspaper3k/
[20] Fuzzy String Matching Using FuzzyWuzzy - HKUST Library, accessed November 4, 2025, https://library.hkust.edu.hk/sc/fuzzywuzzy/
[21] Structured output | Gemini API - Google AI for Developers, accessed November 4, 2025, https://ai.google.dev/gemini-api/docs/structured-output
[22] Prompt Engineering Part 4: Architecting Reliable Outputs with Optimization Techniques | by Akanksha Sinha | Medium, accessed November 4, 2025, https://medium.com/@akankshasinha247/prompt-engineering-part-4-architecting-reliable-outputs-with-optimization-techniques-7bfa0b1b98c7
[23] Deep Learning for Stock Performance Prediction: A Sharpe Ratio-Optimized Approach, accessed November 4, 2025, https://www.researchgate.net/publication/390130909_Deep_Learning_for_Stock_Performance_Prediction_A_Sharpe_Ratio-Optimized_Approach
[24] Can LLM-based Financial Investing Strategies Outperform the Market in Long Run? - arXiv, accessed November 4, 2025, https://arxiv.org/html/2505.07078v1
[25] Stock Price Prediction and Portfolio Optimization Based on Mean Variance Model and Random Forest Model - Atlantis Press, accessed November 4, 2025, https://www.atlantis-press.com/article/126008533.pdf
[26] LLMs for Explainable AI: A Comprehensive Survey - arXiv, accessed November 4, 2025, https://arxiv.org/html/2504.00125v1
How to cite this paper
@article{1717317,
author = {Shreyash Nimbargi, Vedant Joshi, Om Hundekari, Vedant Jagtap},
title = {Deepstock : AI Powered Stock Analysis},
journal = {Iconic Research And Engineering Journals},
year = {2026},
volume = {9},
number = {11},
pages = {753-757},
issn = {2456-8880},
url = {https://www.irejournals.com/formatedpaper/1717317.pdf},
abstract = {The pervasive challenge in modern quantitative finance lies in the failure of traditional models to integrate crucial, non-numerical market factors, leading to poor predictive stability during event-driven volatility. This manuscript details the DeepStock platform, an operative architecture engineered to systematically harmonize structured financial time-series data with highly descriptive, unstructured textual narratives. The system leverages the superior contextual reasoning of the Google Gemini Large Language Model (LLM) to generate robust, quantifiable financial sentiment scores. This sentiment feature vector is fused with traditional technical indicators via an Early Fusion strategy to construct a comprehensive predictive feature set. We hypothesize that models incorporating these hybrid features will yield superior risk-adjusted returns, as validated by the Sharpe Ratio, relative to purely time-series-dependent strategies. The deployment utilizes a Flask/Python framework to provide a free, real-time AI tool focused on analysis and prediction within the NSE and BSE stock markets.},
month = {May},
doi = {https://doi.org/10.64388/IREV9I11-1717317}
}