We grouped the wells by analyzing their surface X and Y coordinates. We used the KMeans algorithm to cluster the wells into 20 groups. Afterwards we used the cluster number to compute the variation with respect to the mean OilPeakRate.
Under the /model directory we store the KMeans model and the cluster labels for each well are stored in data/cluster_label_features.csv.