Seasonal Trend Analysis on Multi-Variate Time Series Data

This research addresses the problem of fluctuation in the hidden patterns within the patterns of financial time series data from the stock market. It becomes even more difficult when the data repository is huge and the data source is from multiple sources. This causes difficulty for the investors to make informed decisions on buying and selling stocks. A novel method is presented to solve this problem using Linear Regression and Random Forest algorithms for multivariate time series data analysis. Initially, data preprocessing is done that includes the data cleaning, integration, transformation, discretization, and data reduction. Then regression analysis is used to predict the dependent variables from independent variables that are known from the past data. The output from the regression analysis is classified using the Random Forest algorithm for learning the patterns. The trend and seasonal information from the patterns will help people to choose better stock market investment.