jkapila / vcvn

R package for Variable Selection, Curve Fitting, Variable Conversion and Normalisation

vcvn

R package for Variable Selection, Curve Fitting, Variable Conversion, Normalisation and Accuracy Measures

This will include following:

Variable Selection:

1. Information Value
2. Gini Index
3. Gini Impurity
4. Entropy Gain
5. Misclassification Error

To be Implemented under Variable Selection Methods:

1. Variable Ranking Methods - voting / scoring / weighted scoring / weighted voting
2. Generic Scoring Function (for Regressiona and Classification)
3. Variable Inflation Factor
4. Other Variable Impacts for regression

Curve Fitting(To be implemented):

1. Template for Curve Fitting for Contineous and Categorical Variable
2. Curve Comparision Methods
3. Curve Indentification
4. Curve Tuning
5. Curve to Normal Conversion
6. Non - Curve / Random / Many Matching curve Decision Criterion
7. Goodness of Fit Test: a) Kolomogorov- Simronov Test b) Carmer-Von Mises Test c) Anderson-Darling Test d) Shapiro -Wilk Test e) Chi-Squared Test f) Akaike Information Criterion (AIC) g) Hosmer - Lemeshow Test

Variable Conversion(To be implemented):

1. Continuous to Categorical a) Range Binning b) WOE Criterion Binning c) Dependent Binning

2. Categrical to Contineous a) One - Hot Encoding with and without reference b) Label Encoding c) Weightage Encoding d) Boosted Encoding ( Based on CatBoost Methodology by Yandex)

Normalisation (To be Implemented):

1. Unit Mean
2. Unit SD
3. Unit Mean And SD
4. Min - Max
5. Box-Cox
6. Log
7. Exponential
8. Mean Difference
9. Median Difference
10. Mean Difference wiht SD
11. Median Difference with SD

**Will also try to include predict function for applying variable conversion and normalisation on raw data.

Measures (These are extensions for other calculations):

1. RMSE
2. MAE
3. MAPE
4. R-squared
5. AIC
6. BIC
7. AUC

To be implemented Measures:

1. Kendall's Tau
2. Gini Index
3. Weights
4. Extension to caret's ConfusionMatrix

**Will also try to include methods for finding best and/or biased limit for probablity cut-off of calssification problem

Updated as on 28th August, 2017