What is Entropy?
Entropy can be defined as a degree of randomness or uncertainty in the system, in terms of Information Entropy, it is the average amount of information contained in a message over entire rage of possibilities of transmitted message.
What is Mutual Information?
It is a way of summarizing much knowing the value of one random variable tells you about another random variable. If X and Y are independent, the mutual information is zero. Any correlations (positive, negative, or nonlinear) will result in positive mutual information.
It is the difference in predictability that you get from knowing the joint distribution p(X,Y) compared to knowing only the two marginal distributions p(X) and p(Y).
Mutual information of X and Y is expressed in terms of entropy (H):
What is Weight of Evidence?
It is a measure of the predictive power of an independent variable in relation to the dependent variable. It is a widely used measure of the “strength” of a grouping for separating events and non-events (default)
WOE = ln (Distribution of Events / Distribution of Non Events)
|Age||Propensity to buy a Motorcycle|
Note: If the variable is not categorical, create bins for the variable
Rules for WOE
- Each category (bin) should have at least 5% of the observations.
- Each category (bin) should be non-zero for both non-events and events.
- The WOE should be distinct for each category. Similar groups should be aggregated.
- The WOE should be monotonic, i.e. either growing or decreasing with the groupings.
- Missing values are binned separately.
Rules for Binning Algorithm
- The WOE should be monotonic i.e. either growing or decreasing with the bins.
- The slope of the Logistic regression with an independent variable having WOE values is not 1 or the intercept is not ln(% of non-events / % of events) then the binning algorithm is wrong.
|Age bins||Count||Events (1’s)||Non-Events ( count - events)||Distribution of Events||Distribution of Non Events||Weight of Evidence|
What is Information Value (IV)?
The Information Value (IV) of a predictor is related to the sum of the (absolute) values for WoE over all groups. It helps to rank variables on the basis of their importance
IV = ∑ (% of non-events - % of events) * WOE
|Age bins||Count||Events (1’s)||Non-Events ( count - events)||Distribution of Events||Distribution of Non Events||WOE||IV|
Information Value Measure:
|Information Value||Variable Predictiveness|
|< 0.02||Not useful for prediction|
|0.02 to 0.1||Weak predictive Power|
|0.1 to 0.3||Medium predictive Power|
|0.3 to 0.5||Strong predictive Power|
|>0.5||Suspicious Predictive Power|