Data Mining Algorithms and Statistical Analysis for Sales Data Forecast

This paper develops and compares different models to forecast new product sales data with increasing sales trend and multiple predictor inputs. In order to analyze new product with increasing sales trend, we developed and evaluated multiple time series forecasting methods, including Exponential Smoothing model, Holt's Linear model, ARMA model, and ARMA wit linear trend models. Furthermore, we created multiple Causal Factor Forecasting models to incorporate various dependent input factors such as sale person's quotes, product pricing, product seasonality factors, to further reduce forecasting error. We analyzed original data regression model, trend and residual regression model, and ARMAV wit linear trend model to consider input factors. We discovered that ARMAV wit linear trend model gives best forecasting accuracy and lowest RSS (Residual Sum of Square). In conclusion, ARMAV with linear trend method is the best benchmark model to forecast sales data for new product with trend and with sales person's inputs.