ABSTRACT
Sales prediction is the process of estimating and analyzing future sales. Sales can be predicted using historical data or based on a set of predictors or pipeline data on leads generated in customer relationship management. Machine learning derived from artificial intelligence (AI) provides algorithms for sales prediction and conversion. Time series ARIMA model is used to predict sales on historic time-stamped data. Residual analysis is used to check the model fit. The model is used to predict sales for the next 12 months. Decision tree algorithm is used to predict sales based on a set of predictor variables. Decision tree is pruned based on hyper parameters and predicted on the test dataset. Predictor variables such as ‘display’ and ‘price’ of the product has an impact on sales in a retail store. Root mean square error is used to validate the model accuracy. Logistic regression algorithm is used to predict whether a lead generated in an organization will be converted into sale or no sale. Confusion matrix provided accuracy of the logistic regression model along with sensitivity and specificity scores. Step by step process for all three algorithms and business insights are explained in this chapter.
