Model Prediction
If the goal is being able to predict accurately on new data - data not used to build the prediction model - then even simple models should be trained and tested in some fashion against data withheld from the model building process. This is true for any kind of model and any kind of data, despite occasional claims to the contrary. The model will also need to be updated periodically because we will have (for example) new customers, some old customers will have left and because - often overlooked - we may be changing the behavior of our data (e.g., customers) by the way we use the model in ways we haven't anticipated.
|Vice President Sales -K.D. Supply Chain Solutions Pvt. Ltd I I Ex Rhenus | Ex CHEP | Ex Godrej | Ex Bisleri | Ex BlowPlast |
8 年excellent start