Thesis 75 p., 18 fig., 30 sources, 9 app.
SCORING, DATA ANALYSIS METHODS, BORROWER CREDIT CAPACITY, MICROFINANCE ORGANIZATION, PYTHON
The paper formulates proposals for the implementation of an automated solution in the field of scoring - assessing the creditworthiness of borrowers in the activities of a microfinance organization. The data processing technologies on which the scoring models are built are considered, a choice is made in favor of models built on machine learning methods. The architecture of the scoring analysis information system is being designed, a software implementation of a machine learning model based on gradient boosting is being developed for assessing the borrower's creditworthiness.
The object of the research is scoring in a microfinance organization.
The purpose of the final qualification work is the design of an information system for scoring analysis for a microfinance organization, development of a software implementation of the scoring analysis model.
Research and development methods and tools: design methods: IDEF0 standards - functional design methodology and DFD - data flow diagrams; universal modeling language UML, BPWin design tools; software implementation tools: Anaconda tooling environment, Python language, libraries of machine learning methods implementation.
Hardware and software requirements: Windows operating system, Linux, Ubunta, etc., virtual environment for Python support, cloud environment for deploying ClickHouse database.
Scope - organizations for assessing the creditworthiness of borrowers.

the archive contains the text of the explanatory note, source files, text programs in Python, a file in mp4 format to demonstrate the work of the program