As we had another episode of extreme rainfall quite close to home last weekend (see here), I remain interested in this topic. Following HansS's suggestion (dank U, Hans!) , I tried to use Decision tree and Linear Regression models on my data. The first results look promising.

For this case, I have defined 'heavy rainfall' as quantities of 50mm or more in 24 hours.

Below is a picture of the workflow :

And below the output of the Scorer node of the Decision tree approach:

The Accuracy is 99,60 percent, But the model is producing many false negatives, not predicting heavy rain when it actually occurs in 301 cases. One of the possible causes is that in my input data heavy rainfall is a 'minority case', so the model tends to classify too many cases into the majority case, no heavy rainfall.

As the whole idea of this exercise is to predict heavy rainfall - the cases we care about, I must try to improve this.

This is an improvement : we have now 'caught' 240 out of 330 cases of heavy rainfall (red box center of screen). The bad side is that we now also get many False Positives (31.257)