Title: | Ensemble Explainable Machine Learning Models |
---|---|
Description: | We introduced a novel ensemble-based explainable machine learning model using Model Confidence Set (MCS) and two stage Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS) algorithm. The model combined the predictive capabilities of different machine-learning models and integrates the interpretability of explainability methods. To develop the proposed algorithm, a two-stage Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS) framework was employed. The package has been developed using the algorithm of Paul et al. (2023) <doi:10.1007/s40009-023-01218-x> and Yeasin and Paul (2024) <doi:10.1007/s11227-023-05542-3>. |
Authors: | Dr. Md Yeasin [aut], Dr. Ranjit Kumar Paul [aut, cre], Dr. Dipanwita Haldar [aut] |
Maintainer: | Dr. Ranjit Kumar Paul <[email protected]> |
License: | GPL-3 |
Version: | 0.1.1 |
Built: | 2025-02-17 03:13:51 UTC |
Source: | https://github.com/cran/EEML |
Ensemble Explainable Machine Learning Models
EEML(df, Weight)
EEML(df, Weight)
df |
List of dataframes containing various explainable scores for each model |
Weight |
Ensemble weights of the models (from weight function) |
ImpScore: Final variable important score of EEML model
Paul, R.K., Das, T. and Yeasin, M., 2023. Ensemble of time series and machine learning model for forecasting volatility in agricultural prices. National Academy Science Letters, 46(3), pp.185-188.
Yeasin, M. and Paul, R.K., 2024. OptiSembleForecasting: optimization-based ensemble forecasting using MCS algorithm and PCA-based error index. The Journal of Supercomputing, 80(2), pp.1568-1597.
library("EEML") df1<- as.data.frame(matrix(rnorm(50) , nrow = 10) ) df2<- as.data.frame(matrix(rnorm(50) , nrow = 10) ) df3<- as.data.frame(matrix(rnorm(50) , nrow = 10) ) rownames(df1)<- rownames(df2)<-rownames(df3)<-paste0("Var", seq(1,10,1)) colnames(df1)<- colnames(df2)<-colnames(df3)<-paste0("Exp", seq(1,5,1)) DF<- list(df1, df2, df3) EEML<-EEML(df=DF,Weight=NULL)
library("EEML") df1<- as.data.frame(matrix(rnorm(50) , nrow = 10) ) df2<- as.data.frame(matrix(rnorm(50) , nrow = 10) ) df3<- as.data.frame(matrix(rnorm(50) , nrow = 10) ) rownames(df1)<- rownames(df2)<-rownames(df3)<-paste0("Var", seq(1,10,1)) colnames(df1)<- colnames(df2)<-colnames(df3)<-paste0("Exp", seq(1,5,1)) DF<- list(df1, df2, df3) EEML<-EEML(df=DF,Weight=NULL)
Selection of Superior Models Using MSC Algorithm
ModelSel(df, Alpha, K)
ModelSel(df, Alpha, K)
df |
Dataframe of predicted values of models with first column as actual values |
Alpha |
Confidence level of MCS tests |
K |
Resampling length |
SelModel: Name of the selected models
Paul, R.K., Das, T. and Yeasin, M., 2023. Ensemble of time series and machine learning model for forecasting volatility in agricultural prices. National Academy Science Letters, 46(3), pp.185-188.
Yeasin, M. and Paul, R.K., 2024. OptiSembleForecasting: optimization-based ensemble forecasting using MCS algorithm and PCA-based error index. The Journal of Supercomputing, 80(2), pp.1568-1597.
Hansen PR, Lunde A, Nason JM, 2011. The model confidence set. Econometrica, 79(2), 453-497
library("EEML") Actual<- as.ts(rnorm(200,100,50)) Model1<- as.ts(rnorm(200,100,50)) Model2<- as.ts(rnorm(200,100,50)) Model3<- as.ts(rnorm(200,100,50)) Model4<- as.ts(rnorm(200,100,50)) Model5<- as.ts(rnorm(200,100,50)) DF <- cbind(Actual, Model1,Model2,Model3,Model4,Model5) SelModel<-ModelSel(df=DF, Alpha=0.2, K=1000)
library("EEML") Actual<- as.ts(rnorm(200,100,50)) Model1<- as.ts(rnorm(200,100,50)) Model2<- as.ts(rnorm(200,100,50)) Model3<- as.ts(rnorm(200,100,50)) Model4<- as.ts(rnorm(200,100,50)) Model5<- as.ts(rnorm(200,100,50)) DF <- cbind(Actual, Model1,Model2,Model3,Model4,Model5) SelModel<-ModelSel(df=DF, Alpha=0.2, K=1000)
Selection of Superior Models Using MSC Algorithm
Weight(ModelSel, Optim = "PSO")
Weight(ModelSel, Optim = "PSO")
ModelSel |
Dataframe of predicted values of selected models with first column as actual values |
Optim |
Optimisation technique |
WeightEn: Ensemble weight of the candidate models
Paul, R.K., Das, T. and Yeasin, M., 2023. Ensemble of time series and machine learning model for forecasting volatility in agricultural prices. National Academy Science Letters, 46(3), pp.185-188.
Yeasin, M. and Paul, R.K., 2024. OptiSembleForecasting: optimization-based ensemble forecasting using MCS algorithm and PCA-based error index. The Journal of Supercomputing, 80(2), pp.1568-1597.
library("EEML") Actual<- as.ts(rnorm(200,100,50)) Model1<- as.ts(rnorm(200,100,50)) Model2<- as.ts(rnorm(200,100,50)) Model3<- as.ts(rnorm(200,100,50)) DF <- cbind(Actual, Model1,Model2,Model3) SelModel<-Weight(ModelSel=DF,Optim="PSO")
library("EEML") Actual<- as.ts(rnorm(200,100,50)) Model1<- as.ts(rnorm(200,100,50)) Model2<- as.ts(rnorm(200,100,50)) Model3<- as.ts(rnorm(200,100,50)) DF <- cbind(Actual, Model1,Model2,Model3) SelModel<-Weight(ModelSel=DF,Optim="PSO")