Summary

Predictive data analysis is the most advanced data analysis method. It largely overlaps with model building and machine learning. In many cases, predictive analysis is a model building. The process of building a model has three steps: predictor selection, model construction, and model validation.

Until a model is fine-tuned and validated, it can be put to use since it may have been under-fitted or over-fitted. Classification is the simplest prediction. The prediction model in classification is called classifier and the consequencer or dependent is called Class label.

Prediction can never be 100 percent accurate because of the unknown. Therefore prediction model can always be improved once a mistake is made or new data becomes available. There is a new concept that emerged in the last few years called “continue learning” or “life-long learning”. It emphasizes the point of model construction is a continuous process.