Scholarlink

The use of decision support systems (DSS) for making decisions in supply chain management is relevant nowadays. Recommender systems (RS) can be used in addition to expert systems during constructing DSS which allow making decisions based on ratings. However, the rating matrix is generally sparse, and the reconstructing process is usually labor-intensive. Using artificial neural networks (ANN) based on associative memory for rating recovery allows to solve the RS efficiency increasing task. An ANN for rating recovery based on Kohonen maps and a one-step training method increases the training speed. An ANN rating recovery based on auto-associative generalized multilayer perceptron and a one-step training method increases learning accuracy. An ANN rating recovery based on a restricted Cauchy machine and a stochastic training method allows users to work with a large matrix of ratings. The numerical research confirmed the performance of the created software and permitted it to be recommended it for use in the RS for recovery ratings.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

The Decision-Making Method Based on a Neural Network Recommender System

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Eugene Fedorov,
  • Anait Karapetyan,
  • Oksana Zelinska,
  • Maryna Leshchenko,
  • Liubov Oksamytna,
  • Olena Kravchenko,
  • Viktoriya Denysenko

摘要

The use of decision support systems (DSS) for making decisions in supply chain management is relevant nowadays. Recommender systems (RS) can be used in addition to expert systems during constructing DSS which allow making decisions based on ratings. However, the rating matrix is generally sparse, and the reconstructing process is usually labor-intensive. Using artificial neural networks (ANN) based on associative memory for rating recovery allows to solve the RS efficiency increasing task. An ANN for rating recovery based on Kohonen maps and a one-step training method increases the training speed. An ANN rating recovery based on auto-associative generalized multilayer perceptron and a one-step training method increases learning accuracy. An ANN rating recovery based on a restricted Cauchy machine and a stochastic training method allows users to work with a large matrix of ratings. The numerical research confirmed the performance of the created software and permitted it to be recommended it for use in the RS for recovery ratings.