Time Series Decision Analysis Based on Least Squares Method and LSTM

This study explores a novel approach combining the least squares technique with LSTM neural networks for time series analysis in decision-making processes. The research framework leverages the precise parameter estimation capabilities of least squares for establishing variable relationships, while utilizing LSTM’s advanced pattern recognition to identify temporal dependencies in sequential data. Through practical implementation, the methodology establishes quantitative relationships between sales performance and pricing strategies using least squares, while employing LSTM architecture for forecasting inventory replenishment needs. The proposed analytical framework demonstrates significant capability in identifying complex temporal patterns, offering empirical evidence for data-informed corporate decision-making. Experimental outcomes substantiate the methodology’s efficacy, presenting innovative tools and insights for temporal data analysis applications.