{"id":3274,"date":"2016-05-23T17:48:27","date_gmt":"2016-05-23T15:48:27","guid":{"rendered":"https:\/\/geekosas.com\/?p=3274"},"modified":"2026-05-23T17:49:05","modified_gmt":"2026-05-23T15:49:05","slug":"segment-customers-step-by-step","status":"publish","type":"post","link":"https:\/\/geekosas.com\/index.php\/2016\/05\/23\/segment-customers-step-by-step\/","title":{"rendered":"Segment customers step by step"},"content":{"rendered":"

Previously I wrote about neural networks (click here to see it)<\/a>. Neural networks and all other "supervised methods<\/a>" are used when you have a sample of values to predict. But when you know what you want to achieve but do not have a sample of the value to predict, the so\u2011called "unsupervised methods" are used.<\/p>\n

A classic problem where this kind of method is applied is customer segmentation<\/a>, where the segments\/groups are not known in advance. Among the methods, one of the most famous is K\u2011Means.<\/p>\n

K\u2011Means is an algorithm used to find groups of individuals with similar characteristics. Similarity or difference is calculated based on the Euclidean distance of their numerical attributes.<\/p>\n

Below we will follow a step\u2011by\u2011step example of how to segment customers in a simplified case.<\/p>\n

<\/p>\n

To make this example, we will generate a random sample of data in two dimensions per case: data traffic (megabytes) and voice traffic (minutes). Each case has been simulated so that it follows a behavior pattern similar to a group.<\/p>\n

With K\u2011Means we will try to find these four groups, which do exist but are unknown to us.<\/p>\n

Below is a plot of the four groups:<\/p>\n

\"\"<\/p>\n

Real Problem:<\/strong><\/p>\n

We are the Data Scientist of a prestigious mobile phone company (personal experience of the writer) and we are tasked with segmenting\/classifying customers (again, nothing far from the writer\u2019s reality). When reviewing the database, we realize that we only have the monthly data and voice consumption of each customer (there could be many more: top\u2011ups, text messages, calls to customer service, etc.).<\/p>\n

Since the scales of data consumption (MB) and voice consumption (minutes) are very different, and K\u2011Means measures similarity by Euclidean distance between cases, we will scale the values between 0 and 1, obtaining the following data:<\/p>\n

\"\"<\/p>\n

Now we are ready to begin. Since K\u2011Means does not determine the number of groups, it is run for different numbers and the results of each case are compared. The choice of the number of groups lies somewhere between mathematics and judgment.<\/p>\n

We will run K\u2011Means from 1 to 6 groups and use both the mathematical method and the "judgment" method to determine the number of groups.<\/p>\n

The first thing to observe is the level of fit for each case. This is the percentage of variance explained by each segmentation, calculated as between_SS \/ total_SS (I won\u2019t go into detail). When segmentation no longer improves significantly as the number of groups increases, it means the new group is similar to an existing one, so they should be the same.<\/p>\n

Below is the fit plot for each case:<\/p>\n

\"\"<\/p>\n

It seems that more than 4 groups does not make sense. Now let\u2019s compare the aggregated values of each group: we will look at their average values and the number of elements in each group.<\/p>\n

Now we will proceed with the non\u2011numerical analysis, which consists of naming the groups. When two groups cannot be differentiated, it means there is no point in separating them (just out of curiosity, we will compare them with the "real" groups).<\/p>\n

1 Group<\/strong><\/p>\n


\n\"\"<\/p>\n\n\n\n\n
group_number<\/th>\nmin<\/th>\nmb<\/th>\nnumber<\/th>\n<\/tr>\n
1) Average of all<\/td>\n71<\/td>\n447<\/td>\n120<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

2 Groups<\/strong><\/p>\n

\"\"
\n<\/p>\n\n\n\n\n\n
group_number<\/th>\nmin<\/th>\nmb<\/th>\nnumber<\/th>\n<\/tr>\n
1) Few Minutes<\/td>\n9<\/td>\n538<\/td>\n60<\/td>\n<\/tr>\n
2) Many Minutes<\/td>\n134<\/td>\n356<\/td>\n60<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

3 Groups<\/strong><\/p>\n

\"\"
\n<\/p>\n\n\n\n\n\n\n
group_number<\/th>\nmin<\/th>\nmb<\/th>\nnumber<\/th>\n<\/tr>\n
1) Low Traffic<\/td>\n13<\/td>\n111<\/td>\n31<\/td>\n<\/tr>\n
2) Many Minutes<\/td>\n134<\/td>\n356<\/td>\n60<\/td>\n<\/tr>\n
3) Many MBs and few Minutes<\/td>\n5<\/td>\n994<\/td>\n29<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

4 Groups<\/strong><\/p>\n

\"\"
\n<\/p>\n\n\n\n\n\n\n\n
group_number<\/th>\nmin<\/th>\nmb<\/th>\nnumber<\/th>\n<\/tr>\n
1) Many MBs<\/td>\n5<\/td>\n994<\/td>\n29<\/td>\n<\/tr>\n
2) Many Minutes<\/td>\n148<\/td>\n23<\/td>\n30<\/td>\n<\/tr>\n
3) Low Traffic<\/td>\n13<\/td>\n111<\/td>\n31<\/td>\n<\/tr>\n
4) High Traffic<\/td>\n120<\/td>\n690<\/td>\n30<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

5 Groups<\/strong><\/p>\n

\"\"
\n<\/p>\n\n\n\n\n\n\n\n\n
group_number<\/th>\nmin<\/th>\nmb<\/th>\nnumber<\/th>\n<\/tr>\n
1) High Traffic<\/td>\n111<\/td>\n699<\/td>\n22<\/td>\n<\/tr>\n
2) High Traffic, especially minutes*<\/td>\n143<\/td>\n666<\/td>\n8<\/td>\n<\/tr>\n
3) Many MBs<\/td>\n5<\/td>\n994<\/td>\n29<\/td>\n<\/tr>\n
4) Many Minutes<\/td>\n148<\/td>\n23<\/td>\n30<\/td>\n<\/tr>\n
5) Low Traffic<\/td>\n13<\/td>\n111<\/td>\n31<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n