Trigger System For Sales Prediction Using Data Mining Techniques

Currently sales prediction have always used a single prediction algorithm. However, it is not possible for a single algorithm to provide sales predictions for all kinds of commodities accurately. Thus selecting a single algorithm for a set of merchandise can be a challenge. In fact, each algorithm produces a different result, and some algorithms can produce more than one type of result. For example, a Decision Trees algorithm can not only be used for prediction, but also as a way to reduce the number of columns in a dataset, because the decision tree can identify columns that do not affect the final mining model. So, a general prediction algorithm for all commodities is needed. Based on the strengths of each type of data mining algorithms, we propose to prepare a trigger system. This trigger system analyses the data sets available for any commodity and after taking the results into account triggers the best algorithm which can be used for predicting the best accurate sales prediction for that particular commodity. The data set for the sales of commodities is freely available on many open source websites like quandl. The output of this project will be an indigenous system which will mine the data and analyse it. After analysis, it will trigger the appropriate algorithm for the sales prediction of the commodity in future time period. This will result in increased accuracy. With an increased accuracy in prediction results obtained from these algorithms, the owner can order the goods according to the predictions made and can thus avoid potential losses. Keywords- Data Mining, Ensemble Learning, Trigger System, Prediction, Svm, K-Means, Naïve Bayes, Decision Tree, C4.5