MUMBAI, India, Aug. 10 -- Intellectual Property India has published a patent application (202541005961 A) filed by Dr S. Brindha on January 24, 2025, for "analyzing Time Series Data Using Attention Network".
Inventors include Dr S. Brindha; F. Bernice Renita; P. Anusuya; S. Dhansri; S. S. Gayathri; and P. S. Sivathimika Sri.
The application for the patent was published on July 31, 2026, under issue no. 31/2026.
Abstract: ABSTRACT OF THE INVENTION ANALYSING TIME SERIES DATA USING ATTENTION NETWORK The present invention proposes an advanced method for analysing stock market time series data using attention-based neural networks. Stock market data, characterized by fluctuations in stock prices, volumes, and other financial indicators over time, is notoriously complex and difficult to predict due to its non-linear behavior, volatility, and long-term ·dependencies. Traditional models, such as ARIMA or linear regression, often fail to capture these intricate patterns and dependencies, leading to suboptimal predictions. This invention addresses these challenges by applying an attention mechanism, which allows the mode! to focus on the most relevant periods within the time series data. The attention mechanism helps the model learn which past price movements or market behaviors are most influential in predicting future price trends. The system incorporates a multi-head attention architecture within an encoder-decoder framework, where the encoder extracts important features from the stock market data, and the decoder generates predictions, detects anomalies, or classifies trends based on the attention weights assigned to different time steps. The attention-based approach provides several advantages for stock market analysis. It can capture long-term dependencies, such as the impact of past market events or economic conditions on future stock prices, which traditional models may overlook. Additionaily, the attention mechanism enables the model to handle noisy or incomplete data by focusing on the most informative data points, making it more robust to market disruptions, missing data, or outliers. This method can be applied to various stock market analysis tasks, including stock price forecasting, detecting unusual market activity (such as price manipulation or anomalies), and identifYing market trends or phases (e.g., bull or bear markets). The attention-based neural network can improve decision-making in trading strategies, portfolio management, and risk assessment by providing more accurate and reliable predictions. Furthermore, the scalable nature of the attention network makes it suitable for analyzing large and high-dimensional stock market datasets, such as minute-by-minute trading data, market indices, and sector performance. Overall, this invention offers a powerful and efficient tool for stock market analysis, significantly improving predictive accuracy and enabling better-informed trading and investment decisions.
Disclaimer: Curated by HT Syndication.
