A web-based application programming that has invloced visual data mining and active learning.
The main idea would be that instead of doing a DM/ML pipeline from scratch without any knowledge on the data by trying an enormous amount of various models and their settings, try to involve the user/expert in the process.
In many application domains, utilizing Machine Learning (ML) and Data Mining (DM) faces on of the following two problems: either, there are not enough data for learning or only a small portion of the large amount of available data is labeled. For such cases, so-called Human-in-the-Loop Machine Learning (HitL-ML) and Visual Data Mining (VDM) approaches are carried out which, in some extent, involve the domain expert in the learning process. The the survey shows the available approaches for HitL-ML and VDM in the literature, provide their comparison and implement a prototype of an Expert-assisted Data Mining/Machine Learning framework in which, instead of doing a DM/ML pipeline from scratch without any knowledge on the data, the user/expert is involved in the process. The work is on the boundary between the research areas of active learning, automated machine learning and data visualization. The web-based application domain for the implemented prototype will be specified based on the outcome of the literature survey.
Survey will be shared once there is a serious applicant. The survery and proposed algorithms are ready, only implementation needed.