Limitations of Predative analytics (4)

 Overfitting and Unfitting Issues 

The predictive model is less accurate due to two fundamental issues which crop up during the process of the model development. The algorithm is being improved continuously, and it causes two types of model failure because the model goes beyond the boundaries of its training data and because it is not able to get enough patterns from the training set. The idea of underfitting and overfitting is especially evident in students who are shown examples of test-preparation and their studying behavior.

The testing model is able to reach accuracy levels that are higher than the best possible by using all the elements of the training data regardless of their impact on beneficial learning. The model hardly shows any error in testing when it is checked with training data but it performs poorly if it is tested with new datasets. A student who just memorizes information without coming up with problem-solving abilities is a similar situation.

For the creation of precise predictive models, learning techniques that can give accurate predictions and at the same time be adaptable to new situations are necessary. The development of models definitely involves numerous rounds of testing with the help of cross-validation methods. The model can be tuned for hours on end since this is the most important phase in creating operational models that deliver the best performance.