CodeFeedr is a Mining Software Repository (MSR) tool designed to efficiently mine massive amounts of streaming data of projects from various sources using Flink’s streaming framework in combination with Kafka. Commissioned by researchers at TU Delft on the field of Data Science and Software Engineering, the goal of this project was to expand further on the product, as it already existed in a development stage. At the start of the project, CodeFeedr consisted of a core pipeline functionality and a limited amount of plugins which process data sources. CodeFeedr-1Up, as this development team calls itself, aimed to achieve two goals: the first goal is increasing the current amount of available plugins, defined by usable software repository sources, to be used by the client; the second goal is to implement a REPL functionality which requests user-friendly SQL-like queries and outputs the queried data stream. Maven, Cargo, NPM and ClearlyDefined have been developed and have extended the CodeFeedr tool. Furthermore, querying on the aforementioned data sources depending on their data structure is possible for sequential pipelines. With user aid and documentation in mind, logical data models of a plugin’s internal structure have been drawn and supplied in the report.

Artificial Intelligence (AI) is increasingly affecting people’s lives. AI is even employed in fields where human lives depend on the AI’s decisions. However, these algorithms lack transparency, i.e. it is unclear how they determine the outcome. If, for instance, the AI’s purpose is to classify an image, the AI will learn this from examples provided to it (e.g. an image of a cow in a meadow). The algorithm can focus on the wrong part of the image. Instead of focusing on the foreground (cow), it could focus on the background (meadow). This way, by focusing on the background, it could produce a false output (e.g. a horse instead of a cow). To show this, an explanation is needed. For this reason, a variety of methods have been created to explain the reasoning behind these algorithms, called explainability methods. In this paper, six local explainability methods are discussed and compared. These methods were chosen as they were the most prominently used approaches for explainability methods for Convolutional Neural Networks (CNN). By comparing methods with analogous characteristics, this paper is going to show what methods exceed others in terms of performance. Furthermore, their advantages and limitations are being discussed. The comparison shows that Local Interpretable Model-agnostic Explanations, Layer-wise Relevance Propagation and Gradient-weighted Class Activation Mapping perform better than Sensitivity Analysis, Deep Taylor Decomposition and Deconvolutional Network, respectively.