gistfile1.txt

###"naive bayesian" approach... if matched in training to test, return the probability that the person
was accepted over all training examples###

#concatenate all base variables except for RESOURCE

train_person = paste(amazon_Xtrain[,2],amazon_Xtrain[,3],amazon_Xtrain[,4],
                     amazon_Xtrain[,5],amazon_Xtrain[,6],amazon_Xtrain[,7],amazon_Xtrain[,8], sep = "")

test_person = paste(amazon_Xtest[,2],amazon_Xtest[,3],amazon_Xtest[,4],
                    amazon_Xtest[,5],amazon_Xtest[,6],amazon_Xtest[,7],amazon_Xtest[,8], sep = "")

#map probability by person

amazon_Xtrain_person_action = as.data.frame(cbind(train_person, amazon_Ytrain))
colnames(amazon_Xtrain_person_action) = c("person", "action")
amazon_Xtrain_person_action$action = as.integer(amazon_Xtrain_person_action$action) - 1 #sets 2 level factor range to [0,1]

probability = aggregate(action ~ person, data = amazon_Xtrain_person_action, mean)

amazon_Xtrain_person_action$probability = probability[,2][match(amazon_Xtrain_person_action[,1], probability[,1])]

#apply prediction to matched entries

nb_predict = numeric()

nb_predict = amazon_Xtrain_person_action$probability[match(test_person,train_person)]

#fill in NA's with best model's predictions

nb_predict[which(!complete.cases(nb_predict))] = best_pred[,11][which(!complete.cases(nb_predict))]

nb_predict = as.matrix(nb_predict)

write.csv(nb_predict, "C:/Users/dylanjf/Desktop/amazon/submission14.csv")