Shvarts M. Hybrid models and methods for predicting recommendations for online store

Українська версія

Thesis for the degree of Candidate of Sciences (CSc)

State registration number

0419U004615

Thesis Registration Form

0419U004615.pdf

Applicant for

Shvarts Mykhailo Ye.

Specialization

01.05.03 - Математичне та програмне забезпечення обчислювальних машин і систем

Date of defense

24-10-2019

Specialized Academic Board

Д 35.052.05

Lviv Polytechnic National University

Essay

In the dissertation, the scientific and practical task of developing hybrid models and methods for predicting recommendations for users of an online store has been solved. The method for calculating vector similarity coefficients for the weighted sum method, which, unlike the existing ones, uses the demographic characteristics of users, improves the accuracy of forecasting recommendations, has been improved. For the first time, based on the concept of using categorical, mixed, and numerical clustering in a single method, a method of searching for user groups has been developed that adapts to the sparseness of the user-subject matrix.

Thesis supervisor

Lobur Mykhaylo V.

Official opponents

Globa Larysa Sergiyivna
Bodnarchuk Ihor Orestovych

Files

autoreferat--aref.doc

dis.doc

Similar theses

0421U104038

Petrenko Maksym V.

Methods of integration of heterogeneous bibliographic data

0421U104040

Artabaev Yurii Z.

A technique for increasing the accuracy of integrated navigation systems based on predictive models using genetic algorithms.

0421U103593

Zubko Roman Anatoliyovych

Improving methods to increase the time efficiency of fractal image compression

0421U103401

Chystyakova Inna S.

Mapping method between the description logic and the relational data model.

0421U103320

Bielozorova Yana Andriivna

Method of application of wavelet analysis in problems of identification of speech information