Or-Based Intelligent Decision Support System for E-Commerce

Ke Zong, Yuan Yuan, Carlos Enrique Montenegro-Marín, Seifedine Nimer KadryJ. Theor. Appl. Electron. Commer. Res 2021 · DOI

TL;DR

This paper introduces an Operational Research-based Intelligent Decision Support System (OR-IDSS) for e-commerce, leveraging fuzzy logic to enhance decision-making in pricing and negotiation. The proposed system demonstrates improved performance, accuracy, and reduced error rates compared to existing methods, offering a valuable tool for optimizing e-commerce operations.

What Problem Does This Paper Address?

In the rapidly evolving landscape of e-commerce, making informed decisions, particularly regarding pricing and negotiation, is crucial for success. Traditional decision support systems (DSS) often struggle with the inherent uncertainties and complexities of human-like reasoning, especially when dealing with subjective factors like customer preferences and market dynamics. Existing e-commerce DSS may collect data but lack the sophisticated analytical capabilities to translate this data into actionable insights for optimal pricing strategies. This paper addresses the need for a more intelligent and adaptive DSS that can effectively handle the nuances of e-commerce transactions, aiming to provide customers with the "right price" through a more sophisticated negotiation process.

Methods

The core of this research lies in the development of an Operational Research-based Intelligent Decision Support System (OR-IDSS) that integrates fuzzy logic theory and machine learning. The system is designed as a hierarchical fuzzy system, allowing for a structured approach to decision-making.

Key methodological components include:

  • Fuzzy Logic: Fuzzy set theory is employed to handle the inherent vagueness and imprecision in e-commerce data and decision-making processes. This allows the system to represent and process information in a way that mimics human reasoning, moving beyond binary (yes/no) logic.
  • Hierarchical Fuzzy System: The system is structured hierarchically, enabling it to break down complex problems into smaller, manageable sub-problems. This approach is particularly useful for complex decision-making scenarios like pricing negotiation.
  • Gradient Descent Learning Algorithm: To train the fuzzy logic system and optimize its parameters, a gradient descent learning algorithm is utilized. This algorithm iteratively adjusts the system's parameters to minimize errors and improve accuracy, adapting to new information and refining its decision-making capabilities.
  • Triangular Membership Functions: For the fuzzy system, triangular membership functions are used. These are computationally efficient and provide a clear way to define the degree to which an input belongs to a fuzzy set.

The paper also introduces a mathematical model to establish a link between influencing factors (Y) and the offered price (x), represented as $x = Q(Y)$. The system then uses fuzzy rules and a gradient descent approach to learn and refine this relationship.

Datasets

The paper does not explicitly mention specific datasets used for training or evaluation. Instead, it refers to experimental results and comparisons with other systems, implying that standard or benchmark datasets for e-commerce decision-making were likely used. The performance is evaluated based on "Total Number of Datasets" and "Available Datasets" in the experimental results section, but the specifics of these datasets are not detailed.

Key Insights and Ideas

  • Fuzzy Logic for E-commerce Nuance: The application of fuzzy logic is a key insight, recognizing that e-commerce decision-making, especially pricing and negotiation, is not always precise. Fuzzy logic allows the system to handle subjective factors and uncertainties more effectively than traditional crisp logic.
  • Hierarchical Structure for Complexity: The hierarchical fuzzy system design is crucial for tackling the multi-faceted nature of e-commerce decision-making. By breaking down the problem, it allows for more focused and accurate analysis at each level.
  • Gradient Descent for Adaptive Learning: The use of gradient descent for learning demonstrates the system's ability to adapt and improve over time. This is vital in dynamic e-commerce environments where market conditions and customer behaviors can change rapidly.
  • Bridging Operational Research and AI: The paper effectively bridges operational research principles with artificial intelligence techniques (fuzzy logic and machine learning) to create a practical and intelligent decision support system.

Key Results

The experimental results indicate that the proposed OR-IDSS system outperforms several existing decision support systems in various aspects:

  • Performance Ratio: The OR-IDSS achieved a performance ratio of 95% with 50 datasets, outperforming systems like B-IDSS (88%), CIDSS (83%), MCGDM (90%), MSDSS (89%), and FARM (90%).
  • Accuracy: With 500 available datasets, the OR-IDSS achieved an accuracy ratio of 9.5%, which is higher than B-IDSS (8.8%), CIDSS (8.3%), MCGDM (9.0%), MSDSS (8.9%), and FARM (9.0%).
  • Error Rate: The paper states that the OR-IDSS has a "very less" error rate compared to other existing methods, as visually represented in Figure 8.
  • Computational Cost: The OR-IDSS demonstrates a low computational cost compared to other existing methods, attributed to the use of triangular membership functions.
  • Precision Ratio: The system achieves a high precision ratio, suggesting its effectiveness in providing accurate negotiation price/discount suggestions.

Limitations and Future Work

  • Dataset Specificity: The paper lacks detailed information about the specific datasets used for training and evaluation, making it difficult to fully assess the generalizability of the results.
  • Computational Overhead of Gradient Descent: While gradient descent is effective for learning, the paper notes that for large datasets, processing one example at a time can lead to significant overhead and a slower learning process compared to batch gradient descent.
  • Future Work Suggestions: The authors suggest that future research should focus on combining customer preference data (from product ratings and social impact) with user engagement and faith matrices to further enhance e-commerce decision-making systems. They also highlight the importance of exploring basic issues and challenges in e-commerce research.

Summary updated

BibTeX
@article{zong2021or,
  author = {Ke Zong and Yuan Yuan and Carlos Enrique Montenegro-Marín and Seifedine Nimer Kadry},
  title = {{Or-Based Intelligent Decision Support System for E-Commerce}},
  journal = {J. Theor. Appl. Electron. Commer. Res.},
  year = {2021},
  doi = {10.3390/jtaer16040065},
}
Figure