Practical Predictive Analytics

Predicting outcomes

SparkR does have a prediction method, aptly named predict(), so we can run predictions on the training set:

• After running the following code, you will observe that the resulting object contains a new column named prediction
• We can also add a unary flag (1) to the results (grp) to indicate that the output is from the training data
• We will also append the total number of rows to each record, since we will need them later for calculations:

        #look at the predictions vs. the training dataset         preds_train <- predict(model, df)         preds_train$grp <- 1         preds_train$totrows = nrow(preds_train) 

The prediction variable is the probability that the outcome of the event (diabetes) will occur:

head(preds_train)