library("tidyverse") options(dplyr.summarise.inform = FALSE) library("class") library("rpart") library("rpart.plot") library("viridis") library("MLmetrics") ## Simulation Parameters: ## simkey {baseline,extended,hint,hintandextended} ## wireless {lora,nbiot} ## wakeupfor {60s,180s} ## seed [1,200] ## node on[0,12] ## isSender {0,1} ## dataSize {1MB} ## Metrics: ## energy [0,+inf) ## nDataRcv [0,+inf) nseed=200 nwakeupfor=2 nwireless=2 nsimkey=4 nsimulations=nseed*nwakeupfor*nwireless*nsimkey # Must be 3200 ## Load data data=suppressMessages(read_csv("../CCGRID2022.csv"))%>%distinct() # Note that in the data experiment wireless=="lora",seed==1,wakeupfor==60,simkey=="baseline" is present 2 times in the CSV file data_seed=data%>%group_by(simkey,wireless,wakeupfor,seed)%>%summarize(energy=sum(energy),coverage=sum(nDataRcv))%>% mutate(efficiency=energy/coverage)%>% ungroup() F1_Score2=function(truth, pred){ result=sapply(c("baseline","extended","hint","hintandextended"),function(c){ cur_truth=truth[truth==c] cur_pred=pred[truth==c] col=paste0("f1_",c) score=F1_Score(cur_truth,cur_pred) if(is.nan(score)){score=0} list(tibble(!!col:=score)) }) do.call("cbind",result) } build_models=function(ignore_hint=TRUE){ ## Prepare data for traning set.seed(1) # Reproducibility wireless_map=c("lora"=1,"nbiot"=2) data_ml=data_seed%>%select(-efficiency,-seed)%>%mutate(wireless=wireless_map[data_seed$wireless]) if(ignore_hint){ data_ml=data_ml%>%filter(simkey!="hint") } train_set=data_ml%>%sample_frac(0.8) # 80% of the data test_set=data_ml%>%suppressMessages(anti_join(train_set)) # 20% of the data ## KNN predict function knn_fn=function(inputs){ as.vector(knn(train=train_set%>%select(-simkey),test=inputs%>%select(-simkey),cl=train_set$simkey,k=10)) } ## Decision tree tree=rpart( simkey ~ wireless + wakeupfor + energy + coverage, data=train_set, method="class", minsplit=60, minbucket=1) ## Tree predict function tree_fn=function(inputs){ as.vector(predict(tree,newdata=inputs%>%select(-simkey),type="class")) } ## Build models models=list(predict_knn=knn_fn,predict_tree=tree_fn) ## Computer performances perfs=sapply(seq(1,20),function(i){ ## Prepare data for traning set.seed(1) # Reproducibility wireless_map=c("lora"=1,"nbiot"=2) data_ml=data_seed%>%select(-efficiency,-seed)%>%mutate(wireless=wireless_map[data_seed$wireless]) if(ignore_hint){ data_ml=data_ml%>%filter(simkey!="hint") } train_set=data_ml%>%sample_frac(0.8) # 80% of the data test_set=data_ml%>%suppressMessages(anti_join(train_set)) # 20% of the data ## KNN knn_predictions=as.vector(knn(train=train_set%>%select(-simkey),test=test_set%>%select(-simkey),cl=train_set$simkey,k=10)) ## Decision tree tree=rpart( simkey ~ wireless + wakeupfor + energy + coverage, data=train_set, method="class", minsplit=60, minbucket=1) tree_predictions=as.vector(predict(tree,newdata=test_set%>%select(-simkey),type="class")) ## Prefs f1_knn=F1_Score2(test_set$simkey,knn_predictions) f1_tree=F1_Score2(test_set$simkey,tree_predictions) accuracy_knn=sum(test_set$simkey==knn_predictions)/length(test_set$simkey) accuracy_tree=sum(test_set$simkey==tree_predictions)/length(test_set$simkey) list(cbind(tibble(model=c("knn","tree")),rbind(f1_knn,f1_tree),tibble(accuracy=c(accuracy_knn,accuracy_tree)))) }) perfs=do.call("rbind",perfs)%>%mutate_if(is.numeric, ~round(.,digits=2)) perfs=perfs%>%group_by(model)%>%summarize( f1_baseline=mean(f1_baseline), f1_hint=mean(f1_hint), f1_extended=mean(f1_extended), f1_hintandextended=mean(f1_hintandextended), accuracy=mean(accuracy)) write.csv(perfs,paste0("figures/f1_scores_offline_ignoreHINT",ignore_hint,".csv"),quote=FALSE,row.names=FALSE) ## Return models models } generate_inputs=function(ignore_hint=FALSE) { ## Prepare data for traning set.seed(1) # Reproducibility wireless_map=c("lora"=1,"nbiot"=2) data_ml=data_seed%>%select(-efficiency,-seed)%>%mutate(wireless=wireless_map[data_seed$wireless]) if(ignore_hint){ data_ml=data_ml%>%filter(simkey!="hint") } train_set=data_ml%>%sample_frac(0.8) # 80% of the data test_set=data_ml%>%anti_join(train_set) # 20% of the data ## KNN training knn_predictions=knn(train=train_set%>%select(-simkey),test=test_set%>%select(-simkey),cl=train_set$simkey,k=10) ## KNN analysis knn_cont_table=table(knn_predictions,test_set$simkey) knn_accuracy=round((sum(diag(knn_cont_table)/sum(rowSums(knn_cont_table))))*100) knn_prop_table=round(prop.table(knn_cont_table),digits=2) ## Decision tree tree=rpart( simkey ~ wireless + wakeupfor + energy + coverage, data=train_set, method="class", minsplit=60, minbucket=1) tree_predictions=predict(tree,newdata=test_set%>%select(-simkey),type="class") tree_cont_table=table(tree_predictions,test_set$simkey) tree_accuracy=(sum(diag(tree_cont_table)/sum(rowSums(tree_cont_table)))) tree_prop_table=round(prop.table(tree_cont_table),digits=2) ## Elbow plot elbow_data=lapply(seq(1,10),function(kvalue){ knn_predictions=knn(train=train_set%>%select(-simkey),test=test_set%>%select(-simkey),cl=train_set$simkey,k=kvalue) ## KNN analysis knn_cont_table=table(knn_predictions,test_set$simkey) knn_accuracy=(sum(diag(knn_cont_table)/sum(rowSums(knn_cont_table)))) knn_prop_table=round(prop.table(knn_cont_table),digits=2) tibble(k=kvalue,accuracy=knn_accuracy) }) elbow_data=do.call("rbind",elbow_data) ggplot(data=elbow_data,aes(k,accuracy))+geom_point()+geom_line()+ggtitle(paste("K-elbow for with NoHint to",as.character(ignore_hint)))+ylim(c(0,1)) ggsave(paste0("figures/knn_elbow_NoHintIs",as.character(ignore_hint),".pdf")) ## Prints print(paste0("Accuracy: KNN=",knn_accuracy,"% CART=",tree_accuracy,"%")) pdf(paste0("figures/tree_",as.character(ignore_hint),".pdf")) tree_plot=rpart.plot(tree,box.palette=as.list(viridis::viridis(4,begin=0.48)),tweak=1.111) silent_call=dev.off() ## Notes: KNN accuracy jump to 76% and CART to 80% accuracy without the hint policy ## Generate simulation inputs inputs=tibble( wakeupfor = c(60,180,60,180), wireless = c("lora", "lora", "nbiot", "nbiot")) constraints=apply(inputs,1,function(row){ wi=row["wireless"] wa=as.numeric(row["wakeupfor"]) ## First extract energy/coverage boundaries min_energy=min((data_seed%>%filter(wireless==wi,wakeupfor==wa))$energy) max_energy=max((data_seed%>%filter(wireless==wi,wakeupfor==wa))$energy) min_coverage=min((data_seed%>%filter(wireless==wi,wakeupfor==wa))$coverage) max_coverage=max((data_seed%>%filter(wireless==wi,wakeupfor==wa))$coverage) ## Generate random points (10 per scenarios) n=100 current_inputs=tibble( wireless=rep(wi,n), wakeupfor=rep(wa,n), energy_constraint=runif(n,min_energy,max_energy), coverage_constraint=round(runif(n,min_coverage,max_coverage))) predictions_knn=knn(train=train_set%>%select(-simkey),test=current_inputs%>% rename(energy=energy_constraint,coverage=coverage_constraint)%>% mutate(wireless=wireless_map[wireless]),cl=train_set$simkey,k=10) predictions_tree=predict(tree,newdata=current_inputs%>% rename(energy=energy_constraint,coverage=coverage_constraint)%>% mutate(wireless=wireless_map[wireless]),type="class") knn_final=tibble(cbind(current_inputs,tibble(simkey=predictions_knn,model="knn"))) tree_final=tibble(cbind(current_inputs,tibble(simkey=predictions_tree,model="tree"))) rbind(knn_final,tree_final) }) inputs=do.call("rbind",constraints)%>%distinct() ## Dimension Energy/Coverage ggplot(data_seed%>%mutate(wakeupfor=as.character(wakeupfor)), aes(coverage,energy,color=simkey))+geom_point()+ geom_point(data=inputs%>%mutate(wakeupfor=as.character(wakeupfor)),aes(coverage_constraint,energy_constraint),size=3,pch=5)+ ggtitle("Dimension Energy/Coverage")+xlab("Coverage")+ylab("Sum of nodes energy consumption (J)")+ facet_wrap(~wakeupfor+wireless,scale="free") ggsave(paste0("figures/random_inputs_NoHintIs",as.character(ignore_hint),".pdf"),width=15) write.csv(inputs,paste0("inputs/inputs_NoHintIs",as.character(ignore_hint),".csv"),row.names=FALSE, quote=FALSE) } ## Build models and generate performance metrics build_models(ignore_hint=FALSE) build_models(ignore_hint=TRUE) ## Generate inputs generate_inputs(ignore_hint=FALSE) generate_inputs(ignore_hint=TRUE)