Method for classifying user action sequence

A method for training a Machine Learning Algorithm (MLA), the MLA executed by a server, the MLA for classifying a user action sequence that is performed by a user with an electronic service using a computer device, the method executable by the server, the method comprising:

• receiving an indication of interface elements of the electronic service and events associated with the interface elements to be monitored;
• receiving a plurality of indications of the user action sequence, the user action sequence including occurrence of at least one of:
• (i) events associated with the interface elements and (ii) user interactions with the interface elements;
• and associated timestamps,
• the plurality of indications being of at least two different types of classes, for which the MLA is to be trained for classifying user actions into;
• generating a training set comprising user action sequences belonging to the at least two different types of classes;
• the generating a given training set having a given user action sequence including:
• subdividing the given user action sequence into subsequences, the subdividing being based on a pre-determined set of:
• max subsequence length,
• min subsequence length;
• determining a frequency of each subsequence appearing in the user action sequences belonging to a given one of the at least two different types of classes;
• scoring each subsequence based on the frequency;
• selecting n most informative subsequences indicative at the probability of the associated given user action sequence belonging to the given one of the at least two different types of classes;
• using the training set to train the MLA to classify an in-use user action sequence into one of the at least two types of classes.