Devcon Challenge

Results for liblinear using a full set of features (noscale) ( 12 - L2-regularized L2-loss support vector regression (dual))

#readr::write_csv(RESULTS,path = "./devcon_challenge.liblinear_12_results_noscale_all_lines.csv")
a0<-readr::read_csv("./devcon_challenge.liblinear_12_results_noscale_all_lines.csv") %>% filter(!is.na(pearson)) 
a0

readr::read_csv("./devcon_challenge.liblinear_12_results_noscale_all_lines.csv") %>% filter(!is.na(pearson)) %>%  summarise(pearson_mean=mean(pearson),spearman_mean=mean(spearman))

Results for liblinear (noscale) 1500 features. ( 12 - L2-regularized L2-loss support vector regression (dual))

#readr::write_csv(RESULTS,path = "./devcon_challenge.liblinear_12_results_noscale_rf_selected_features_1500_all_lines.csv")
a6<-readr::read_csv("./devcon_challenge.liblinear_12_results_noscale_rf_selected_features_1500_all_lines.csv") %>% filter(!is.na(pearson)) 
a6

readr::read_csv("./devcon_challenge.liblinear_12_results_noscale_rf_selected_features_1500_all_lines.csv") %>% filter(!is.na(pearson))  %>%  summarise(pearson_mean=mean(pearson),spearman_mean=mean(spearman))

Results for liblinear (noscale) 1000 features. (13 L2-regularized L1-loss support vector regression (dual))

#readr::write_csv(RESULTS,path = "./devcon_challenge.liblinear_13_results_noscale_rf_selected_features_1000_all_lines.csv")
a1<-readr::read_csv("./devcon_challenge.liblinear_13_results_noscale_rf_selected_features_1000_all_lines.csv") %>% filter(!is.na(pearson)) 
a1

readr::read_csv("./devcon_challenge.liblinear_13_results_noscale_rf_selected_features_1000_all_lines.csv") %>% filter(!is.na(pearson)) %>%  summarise(pearson_mean=mean(pearson),spearman_mean=mean(spearman))

Results for liblinear (noscale) 1500 features. (13 L2-regularized L1-loss support vector regression (dual))

#readr::write_csv(RESULTS,path = "./devcon_challenge.liblinear_13_results_noscale_rf_selected_features_1500_all_lines.csv")

a2<-readr::read_csv("./devcon_challenge.liblinear_13_results_noscale_rf_selected_features_1500_all_lines.csv") %>% filter(!is.na(pearson)) 
a2

readr::read_csv("./devcon_challenge.liblinear_13_results_noscale_rf_selected_features_1500_all_lines.csv") %>% filter(!is.na(pearson))  %>% summarise(pearson_mean=mean(pearson),spearman_mean=mean(spearman))

NA

Results for liblinear (noscale) 2000 features. (13 L2-regularized L1-loss support vector regression (dual))

#readr::write_csv(RESULTS,path = "./devcon_challenge.liblinear_13_results_noscale_rf_selected_features_2000_all_lines.csv")

a3<-readr::read_csv("./devcon_challenge.liblinear_13_results_noscale_rf_selected_features_2000_all_lines.csv") 
a3

readr::read_csv("./devcon_challenge.liblinear_13_results_noscale_rf_selected_features_2000_all_lines.csv")  %>%  summarise(pearson_mean=mean(pearson),spearman_mean=mean(spearman))

Results for glmnet using 5000 features

#readr::write_csv(RESULTS,path = "./devcon_challenge.gmlnet_results_noscale_rf_selected_features_5000_all_lines.csv")
a4<-readr::read_csv("./devcon_challenge.gmlnet_results_noscale_rf_selected_features_5000_all_lines.csv") %>% filter(!is.na(pearson)) 
a4

readr::read_csv("./devcon_challenge.gmlnet_results_noscale_rf_selected_features_5000_all_lines.csv") %>% filter(!is.na(pearson)) %>%  summarise(pearson_mean=mean(pearson),spearman_mean=mean(spearman))

Results for glmnet using 500 features

#readr::write_csv(RESULTS,path = "./devcon_challenge.gmlnet_results_noscale_rf_selected_features_500_all_lines.csv")
a5<-readr::read_csv("./devcon_challenge.gmlnet_results_noscale_rf_selected_features_500_all_lines.csv") 
a5

readr::read_csv("./devcon_challenge.gmlnet_results_noscale_rf_selected_features_500_all_lines.csv") %>% filter(!is.na(pearson)) %>%  summarise(pearson_mean=mean(pearson),spearman_mean=mean(spearman))

library(gridExtra)
a0$type<-"liblinear12full"
a1$type<-"liblinear13_1000"
a2$type<-"liblinear13_1500"
a3$type<-"liblinear13_2000"
a4$type<-"glmnet5000"
a5$type<-"glmnet500"
a6$type<-"liblinear12_1500"

plot2<-rbind(a0,a1,a2,a3,a4,a5,a6) %>% group_by(type,datasets) %>% summarise(pearson=mean(pearson)) %>% arrange(pearson) %>%
  ggplot()+
  geom_tile(aes(x=type,y=datasets,fill=pearson))+
  scale_fill_gradient(low = "red", high = "white",limits=c(0.4, 1),)+
  theme(axis.text.x=element_text(angle=45,hjust=1))
  
  
plot1<-rbind(a0,a1,a2,a3,a4,a5,a6) %>% group_by(type,datasets) %>% summarise(spearman=mean(spearman)) %>% arrange(spearman) %>%
  ggplot()+
  geom_tile(aes(x=type,y=datasets,fill=spearman))+
  scale_fill_gradient(low = "red", high = "white",limits=c(0.4, 1),)+
  theme(axis.text.x=element_text(angle=45,hjust=1))

gridExtra::grid.arrange(plot1,plot2,ncol=2)  

NA