A method of evaluating and selecting an application programming interface (API). The method includes accessing source-code in computer programming languages from an API repository. The method includes training a machine learning classifier using generalized, extracted API-calls to learn variations of API-call programming language syntaxes through usage of the API-call programming language syntaxes in the source-code. The method includes building API-call models suitable for the computer programming languages and building a parameter relations graph based on backtracking data-flow from API-calls that are extracted from public project repositories based on the API-call model. The method includes retrieving sample parameter values from the parameter relations graph. The method includes implementing the sample parameter values in an evaluation interface to evaluate functionality of an API in a software application.