Techniques for approximate neighbor selection in tree models for computing co-occurrence

Abstract

Certain example embodiments relate to techniques for building a contextual recommendation engine. Flow representations indicating how services are to interact with one another are received. Each service represents a program logic segment. Flow representations are parsed to identify invocation and conditional processing operations therein. These identified operations are arranged in a predefined format including sequences. Each sequence corresponds to a respective flow representation and includes any identified operations in an order matching their occurrences in the respective flow representation. For each identified conditional processing operation, each identified invocation operation associated with a branch thereof is merged into a single path for inclusion in the respective sequence in place of the respective identified conditional processing operation. Each sequence is converted into a listing of atomic units. A co-occurrence model based on the listing is developed and stored to a model store so that it can be queried as integrations are built.