Stock market forecasting is not an easy task to undertake because of the volatility and the price movements which are non-stationary. In this work, the author suggests the hybrid deep learning architecture that combines the use of Discrete Wavelet Transform (DWT)-based feature extraction with multi-architecture aggregation with LSTM, GRU, RNN, and CNN models. The model was tested on six Indian stocks based on ten-years of daily historical returns under a rolling walk-forward validation procedure, with the model being trained on five-year window and tested on out of sample periods. It has been shown experimentally that the hybrid aggregation method gives smaller prediction errors than individual architectures. Using the proposed model on the NIFTY 50 index, the RMSE was 0.2102, and the directional accuracy was 65.18, which is a better predictive stability. An ablation analysis also supports the fact that wavelet-based pre-processing helps to reduce errors and increase consistency of trends. These results indicate that wavelet-based feature extraction with heterogeneous deep learning models can be more robust when used in daily stock forecasting. In the future, statistical significance test and trading simulations based on costs may be introduced to the work to conduct additional testing of their practical relevance.
Priya Sidhu, Himanshu Aggarwal, Madan Lal· International Journal of Int...· 0 citations
The evolution of approaches to predicting trends in financial markets has involved several stages, starting from autoregressive statistics, going further to deep sequences, graphs, fusion of sentiment with LLMs, and finally arriving at large language model (LLM) reasoning layers. One area where there has been a lack is the gap between prediction and recommendation, however, because features such as transaction costs, diversification, and investor risk profile are not taken into consideration, a prediction model may indicate a reasonable directionality for held-out samples but be uneconomical to apply when including those factors. This paper presents a survey of the current state of the art for financial trend prediction for the 2025–2026 horizon and eight methodological families, including the classical machine learning methods, recurrent sequences, attention and transformers, graph neural networks, sentiment and LLM-based fusion, explainable AI, reinforcement-learning-based trading agents, and recommender-system-specific approaches. The review shows that predictive accuracy gains have become incremental with respect to the complexity of the architecture needed to achieve them and that the evaluation methodology is now only just catching up with what is important economically and that the number of systems for which the explicit intent is that of a recommendation is still a minority compared to those that are only forecast. In this space, we motivate a design framework for a stock recommender system where prediction is not an end goal, but is one of many inputs, and we outline the open challenges which such a system must still address before it can be trusted with real money.
Priya Sidhu, H. Aggarwal, Madan Lal· Journal of Intelligent Decis...· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
