Logo image
Back
Enhancing Stock Market Predictions through CNN-Based Hybrid Deep Learning Models
Conference proceeding

Enhancing Stock Market Predictions through CNN-Based Hybrid Deep Learning Models

Suresh Talamala, Ushasree R, F. Praveena, V. Manoranjithem, Saginala Mastan and Anvesh Perada
2025 5th Asian Conference on Innovation in Technology (ASIANCON), pp 1-6
22 Aug 2025
DOI: https://doi.org/10.1109/ASIANCON66527.2025.11280763

Abstract

Accuracy Data models Efficient Market Hypothesis (EMH) Ensemble Adaptive Neuro-Fuzzy Inference System (ENANFIS) Feature extraction Indexes Investment Oscillators Predictive models Relative Strength Index (RSI) Stochastic processes Stock Market Predication (SMP) Stock markets Training
Due of the huge stakes and ever-changing nature of market behaviour, mathematical, engineering, and financial analysts have taken an intense interest in stock market prediction. The stock market is the pinnacle of investment venues due to the massive amount of money that flows through it. But, even with the EMH to consider, which states that markets are intrinsically efficient and unpredictable, it is still a difficult undertaking to predict market movements exactly. Predictive modelling is now within reach of ordinary and institutional investors alike thanks to the proliferation of financial data and the improvement in processing power. An improved model for stock market prediction using a hybrid LSTCN is suggested in this paper. First, the data is preprocessed to remove outliers and missing values. Then, technical indicators like the RSI and stochastic oscillators are extracted. To handle the unpredictability of financial data, the model uses quantile regression and an attention mechanism. The experimental results demonstrate that the hybrid LSTCN model achieves a remarkable 94.63% accuracy, surpassing that of conventional models. This proves that the model can accurately predict future outcomes. When it comes to making better stock market investments, the suggested method shows a lot of promise.

Metrics

1 Record Views

Details

Logo image