Scratchpad for GovHack 2017 entry - Project Blue Orange

admin_govhack_ml.R

# Load libraries.
library(ggplot2);
library(caret);
library(gridExtra);

set.seed(100);

# Load our data.
child <- read.csv("childcare_dataset/childcare_dataset.csv", stringsAsFactors = FALSE);
head(child);

child$sa3_code <- as.numeric(child$sa3_code);

# Calculate the state from the SA3 code.
child$state <- ifelse(child$sa3_code >= 20000, "VIC", ifelse(child$sa3_code >= 30000, "QLD", ifelse(child$sa3_code >= 40000, "SA", "NSW")));

child$child_age <- as.factor(child$child_age);
child$service_type <- as.factor(child$service_type);
child$state <- as.factor(child$state);
child$top_industry <- as.factor(child$top_industry);

child$sa3_code <- NULL;
child$sa3_name <- NULL;
child$child_count <- NULL;

# Split our data into a training (70%) and test (30%) set.
training_split = createDataPartition(y = child$service_count_band, p = 0.70, list = FALSE);
training_set = child[training_split,];
testing_set  = child[-training_split,];

# Train our model.
#model_fit = train(service_count_band ~ ., method = "rpart", data = training_set, trControl = trainControl(method = 'cv', number = 10, classProbs = TRUE)); # Using all features.

#model_fit = train(service_count_band ~ ., method = "knn", data = training_set, trControl = trainControl(method = 'cv', number = 10, classProbs = TRUE)); # Using all features.

model_fit = train(service_count_band ~ ., method = "rf", data = training_set, trControl = trainControl(method = 'cv', number = 10, classProbs = TRUE)); # Using all features.

# Use model to predict the service count bands.
testing_set_predict = predict(model_fit, newdata = testing_set[, -3]);

# Output confusion matrix
table(testing_set_predict, testing_set$service_count_band);

testing_set$Correct = testing_set_predict == testing_set$service_count_band;
accuracy = length(testing_set$Correct[testing_set$Correct == TRUE]) / length(testing_set$Correct);
paste("Training accuracy:", accuracy);

preProcess(training_set[, -3], method = "pca", thresh = 0.99);

testing_set_predict = predict(model_fit, newdata = testing_set[, -3], "prob");

# Inspect the most important features.
varImp(model_fit);

childcare_dataset.csv