
290 Empirical Research in Software Engineering
7.4 Concerns in Model Prediction
The prediction models must be carefully developed. Before and during model prediction,
there are many issues and concerns that must be addressed.
• Data must be preprocessed using outlier analysis, normality tests and so on. It
may help in increasing the accuracy of the models. Section 6.1presents the pre-
processing techniques.
• The model must be checked for multicollinearity effects (see Section 7.4.2).
• Dealing with imbalanced data (see Section 7.4.3).
• A suitable learning technique must be selected for model development (see
Section7.4.4).
• The training a ...