Apache Spark Deep Learning Cookbook

There's more...

We now know that our model was able to accurately predict whether a call coming in is related to fire or not at a rate of 88.4 percent. At first, this may sound like a strong prediction; however, it's always important to compare this to a baseline score where every call was predicted as a non-fire call. The predicted dataframe, df_predicted, had the following breakdown of labels 1 and 0, as seen in the following screenshot:

We can run some statistics on that same dataframe to get the mean of label occurrences of value 1 using the df_predicted.describe('label').show() script. The output of that script can be seen in the following ...