Abstract:
Almost every industry, organizations use Information Technology for their core
business activities. Similarly, SMEs tend to use technologies for their businesses and
the usage has increased data generation. Even though the amount of data generated
low compared to the Large-Scale Enterprises, yet SMEs’ data reserves can be
converted into meaningful information with the aid of data mining and machine
learning concepts. When using data mining and machine learning approaches for
datasets generated at SMEs need special cautions, due to constraints such as limited
computational capacity at SMEs and as well as comparatively less data capacity at
data repository than convectional Big data. This study aims to formulate a data
mining model for sales prediction and as well as focus on initial Customer
Relationship Management stage, customer identification within the context of
SMEs, which is effective and efficient under the constraints identified. The data sets
comprise of demographic features; Gender, Age, Residency information, Marital
Status and Occupation. Preprocessed repository data used to initiate descriptive
statistical analysis, variable correlation analysis and feature engineering. The
processed variables used within number of machine learning algorithms; Linear
Regression (LR), Ridge Regression (RR), Decision Tree Regression (DTR),
Random Forest Regression (RFR), Multiple Layer Perceptron Regression (MLPR)
models to find most efficient algorithm for sales prognostication. Furthermore, the
customer identification had initiated with Principle Component Analysis (PCA)
along with classification techniques. The finalized model for predictive analytics at
SMEs determined through the lowest Root Mean Square Error (RMSE) and then a
validation process carried out to assess the performance. For the predictive analytics
in sales, DTR was suitable due to lowest RMSE of 2689 and through PCA and
classification 3 customer bases were identified.
