#QSAR su intrinciniais parametrais
#
# VR

#pradiniai duomenys:
deskriptoriai<-as.matrix(read.csv(file='sutvarkyti_descriptoriai.csv'))
matavimai<-read.csv(file='Intinciniai_matavimai.csv')

#random numbers:
# test <- sort(round(runif(8, 1, 28)))
# test
#test <- c(1, 9, 10, 13, 14, 19, 22, 26)
#tada:
#train <- c(2, 3, 4, 5, 6, 7, 8, 11, 12, 15, 16, 17, 18, 20, 21, 23, 24, 25, 27, 28)

# gal geriau tiesiog ismesti 3 nefuorintus
# test <- sort(round(runif(5, 1, 25)))
test <- c(8, 11, 13, 17, 23)
train <- c(1, 2, 3, 4, 5, 6, 7, 9, 10, 12, 14, 15, 16, 18, 19, 20, 21, 22, 24, 25)


library(cvq2)

#Q2TEST f-ja kaip ir PHASE Q2 lygiai tokia pati
q2test<-function(activity, predicted_activity) {
                        prediction_error_sq<-(predicted_activity-activity)^2
                avg_activity<-mean(activity)
                                sigma_y_sq<-(activity-avg_activity)^2
                                q2test_val<-1-sum(prediction_error_sq)/sum(sigma_y_sq)
                                                return(q2test_val)
}



#Genetic Algorithm (GA) for Variable Selection from High-Dimensional Data:
library(gaselect)
ctrl <- genAlgControl(populationSize = 5000, numGenerations = 750, minVariables = 3,
                      maxVariables = 4, verbosity = 1)
evaluatorRDCV <- evaluatorPLS(numReplications = 2, innerSegments = 5, outerSegments = 3,
                              numThreads = 3)

#GintrCAI:
#cia su greitesniu nustatymu buvo...:"WD.unity", "E1v","maxssCH2", "TDB5u"
#resultRDCV.GintrCAI <- genAlg(matavimai$GintrCAI[train], deskriptoriai[train,], control = ctrl, evaluator = evaluatorRDCV, seed = 123)
#subsets(resultRDCV.GintrCAI, 1:5)

qsar_1_train<-lm(matavimai$GintrCAI[train] ~ deskriptoriai[train, "WD.unity"] + deskriptoriai[train, "E1v"] + deskriptoriai[train, "maxssCH2"] + deskriptoriai[train, "TDB5u"])
print(summary(qsar_1_train))
qsar_1_test_pred_values<-coef(qsar_1_train)[1] + coef(qsar_1_train)[2]*deskriptoriai[test, "WD.unity"] + coef(qsar_1_train)[3]*deskriptoriai[test, "E1v"] + coef(qsar_1_train)[4]*deskriptoriai[test, "maxssCH2"] + coef(qsar_1_train)[5]*deskriptoriai[test, "TDB5u"]
qsar_1_test<-lm(qsar_1_test_pred_values ~ matavimai$GintrCAI[test])
print(summary(qsar_1_test))
x<-cbind(deskriptoriai[train, "WD.unity"], deskriptoriai[train, "E1v"], deskriptoriai[train,"maxssCH2"], deskriptoriai[train, "TDB5u"], matavimai$GintrCAI[train])
colnames(x)<-c('x1', 'x2', 'x3', 'x4', 'y')
qsar_1_q2<-cvq2(x)
print(qsar_1_q2)
print(q2test(matavimai$GintrCAI[test], qsar_1_test_pred_values))

#HintrCAI:
#resultRDCV.HintrCAI <- genAlg(matavimai$HintrCAI[train], deskriptoriai[train,], control = ctrl, evaluator = evaluatorRDCV, seed = 123)
#subsets(resultRDCV.HintrCAI, 1:5)

qsar_2_train<-lm(matavimai$HintrCAI[train] ~ deskriptoriai[train, "Wlambda2.unity"] + deskriptoriai[train, "S9"] + deskriptoriai[train, "MATSm4"] + deskriptoriai[train, "MoRSEV15"])
print(summary(qsar_2_train))
qsar_2_test_pred_values<-coef(qsar_2_train)[1] + coef(qsar_2_train)[2]*deskriptoriai[test, "Wlambda2.unity"] + coef(qsar_2_train)[3]*deskriptoriai[test, "S9"] + coef(qsar_2_train)[4]*deskriptoriai[test, "MATSm4"] + coef(qsar_2_train)[5]*deskriptoriai[test, "MoRSEV15"]
qsar_2_test<-lm(qsar_2_test_pred_values ~ matavimai$HintrCAI[test])
print(summary(qsar_2_test))
x<-cbind(deskriptoriai[train, "Wlambda2.unity"], deskriptoriai[train, "S9"], deskriptoriai[train,"MATSm4"], deskriptoriai[train, "MoRSEV15"], matavimai$HintrCAI[train])
colnames(x)<-c('x1', 'x2', 'x3', 'x4', 'y')
qsar_2_q2<-cvq2(x)
print(qsar_2_q2)
print(q2test(matavimai$HintrCAI[test], qsar_2_test_pred_values))

#TsintrCAI:
#dar ilgesni skaiciavimai
ctrl <- genAlgControl(populationSize = 50000, numGenerations = 750, minVariables = 3,
                      maxVariables = 4, verbosity = 1)

#resultRDCV.TsintrCAI <- genAlg(matavimai$TsintrCAI[train], deskriptoriai[train,], control = ctrl, evaluator = evaluatorRDCV, seed = 123)
#subsets(resultRDCV.TsintrCAI, 1:5)

qsar_3_train<-lm(matavimai$TsintrCAI[train] ~ deskriptoriai[train, "MOMI.Z"] + deskriptoriai[train, "S9"] + deskriptoriai[train, "MoRSEV15"] + deskriptoriai[train, "hmin"])
print(summary(qsar_3_train))
qsar_3_test_pred_values<-coef(qsar_3_train)[1] + coef(qsar_3_train)[2]*deskriptoriai[test, "MOMI.Z"] + coef(qsar_3_train)[3]*deskriptoriai[test, "S9"] + coef(qsar_3_train)[4]*deskriptoriai[test, "MoRSEV15"] + coef(qsar_3_train)[5]*deskriptoriai[test, "hmin"]
qsar_3_test<-lm(qsar_3_test_pred_values ~ matavimai$TsintrCAI[test])
print(summary(qsar_3_test))
x<-cbind(deskriptoriai[train, "MOMI.Z"], deskriptoriai[train, "S9"], deskriptoriai[train,"MoRSEV15"], deskriptoriai[train, "hmin"], matavimai$TsintrCAI[train])
colnames(x)<-c('x1', 'x2', 'x3', 'x4', 'y')
qsar_3_q2<-cvq2(x)
print(qsar_3_q2)
print(q2test(matavimai$TsintrCAI[test], qsar_3_test_pred_values))

#GintrCAII:
#dar kitokie nustatymai:
ctrl <- genAlgControl(populationSize = 50000, numGenerations = 1000, minVariables = 2, maxVariables = 3, verbosity = 1)
#Tikriausiai, kad neoverfitint tai imti maziau deskriptoriu, jei jau su tiek iseina tai ok matyt.
#resultRDCV.GintrCAII <- genAlg(matavimai$GintrCAII[train], deskriptoriai[train,], control = ctrl, evaluator = evaluatorRDCV, seed = 123)
#subsets(resultRDCV.GintrCAII, 1:5)

qsar_4_train<-lm(matavimai$GintrCAII[train] ~ deskriptoriai[train, "RPCS"] + deskriptoriai[train, "nRotBt"] + deskriptoriai[train, "RDF55s"])
print(summary(qsar_4_train))
qsar_4_test_pred_values<-coef(qsar_4_train)[1] + coef(qsar_4_train)[2]*deskriptoriai[test, "RPCS"] + coef(qsar_4_train)[3]*deskriptoriai[test, "nRotBt"] + coef(qsar_4_train)[4]*deskriptoriai[test, "RDF55s"]
qsar_4_test<-lm(qsar_4_test_pred_values ~ matavimai$GintrCAII[test])
print(summary(qsar_4_test))
x<-cbind(deskriptoriai[train, "RPCS"], deskriptoriai[train, "nRotBt"], deskriptoriai[train,"RDF55s"], matavimai$GintrCAII[train])
colnames(x)<-c('x1', 'x2', 'x3', 'y')
qsar_4_q2<-cvq2(x)
print(qsar_4_q2)
print(q2test(matavimai$GintrCAII[test], qsar_4_test_pred_values))