A Hybrid ANN Random Forest Regression Method of Feature Extraction in Time Series Data

A time series is a logical grouping of values about time gleaned from numerous applications. The fundamental properties of the time series data (TSD) include their enormous quantity, high complexity, and traits like trend, cycle, seasonality, and irregularity. Data mining has seen a significant amount of research and development efforts due to the cumulative use of TSD. The effectiveness of various methods of data mining used for time series analysis is closely correlated to the dimension of the time series. In this proposed method, the weather forecasting dataset has been used. The dataset has more number of dimensions and it has been reduce dimensionality using Artificial Neural Network (ANN) and extract the feature with accuracy using Random Forest regression algorithm The study aims to utilize the deep learning (DL) method and feature extraction to minimize the dimensionality in time series data. The proposed approach improves the extraction's effectiveness in reaching our research's goal. The proposal seeks to minimize time series data's high dimensionality. The idea also emphasizes the importance of accurately lowering the dimensionality of time series data. The proposed Random Forest regression method with MAPE (13%), RMSE (9%), and MSE (15%) effectively decreases the dimensionality and extract the feature effectively when compared to existing techniques.