Credit Evaluation System Based on Blockchain for Multiple Stakeholders in the Food Supply Chain

Dianhui Mao, Fan Wang, Zhihao Hao, Haisheng LiInternational Journal of Environmental Research and Public Health 2018 · DOI

TL;DR

This paper introduces a novel credit evaluation system for the food supply chain that leverages blockchain technology and a deep learning model (LSTM) to address information asymmetry and enhance regulatory oversight. By storing transaction and credit evaluation data on a blockchain and analyzing sentiment from textual reviews using LSTM, the system aims to create a more transparent, accountable, and trustworthy food supply chain ecosystem.

What Problem Does This Paper Address?

The food supply chain is a complex network involving numerous stakeholders, from farmers to retailers. A significant challenge within this chain is "information asymmetry," where participants may selectively share information to their advantage, leading to issues like food fraud and safety concerns. Existing food safety systems often focus on traceability but lack robust mechanisms for supervising and managing the behavior of these diverse stakeholders. Regulators struggle to obtain reliable and authentic information for effective oversight. This paper tackles the problem of enhancing supervision and management in the food supply chain by establishing a credit evaluation system that promotes accountability and transparency among multiple participants.

Methods

The core of the proposed system is a hybrid approach combining blockchain technology and a deep learning model (LSTM).

  • Blockchain (Hyperledger Fabric): This forms the backbone of the system, providing a decentralized, immutable, and transparent ledger for storing transaction and credit evaluation data. Smart contracts, implemented as "chaincode" in Go, automate the processes of transaction completion and credit evaluation. Hyperledger Fabric is chosen for its consortium nature, allowing for permissioned access and differentiated roles for traders and regulators, while also offering privacy for participants.
  • Long Short-Term Memory (LSTM) Network: This deep learning model is employed for sentiment analysis of textual credit evaluations provided by traders. Unlike traditional methods that might rely on simple "praise" or "bad review" classifications, LSTM directly analyzes the sentiment expressed in free-text reviews. This allows for a more nuanced understanding of trader performance and helps overcome the limitations of simpler credit scoring models that don't account for the sentiment of the evaluation text itself.

The system's workflow involves traders completing transactions and providing credit evaluations via smart contracts. This data is recorded on the blockchain. Regulators can then access this data and use the trained LSTM model to analyze the sentiment of the credit evaluations, generating a credit score that aids in their supervision and management efforts.

Datasets

The paper utilizes a Chinese text dataset of reviews for experimental purposes. This dataset was collected by an internet user named Jianlin Su from an online source (https://spaces.ac.cn/archives/3414). The crawler code was written in Python, and the corpus included over 20,000 reviews, with a specific focus on reviews related to "food." The dataset was pre-processed, including word segmentation using NLPIR-ICTCLAS2016 and vectorization using word2vec. For training and validation, the dataset was randomly divided into a 7:3 ratio.

Key Insights and Ideas

  • Blockchain for Accountability, Not Just Traceability: While blockchain is often discussed for food traceability, this paper highlights its potential for enhancing accountability and supervision by creating a tamper-resistant record of trader interactions and evaluations.
  • LSTM for Nuanced Sentiment Analysis: Moving beyond simple positive/negative ratings, using LSTM to analyze the sentiment of free-text reviews provides a richer and more accurate assessment of trader performance. This is crucial for capturing subtle feedback that might be missed by simpler models.
  • Addressing Information Asymmetry: By making transaction and evaluation data transparent and immutable on the blockchain, the system directly combats information asymmetry, forcing participants to be more accountable for their actions.
  • Dual Role of Blockchain: The system effectively uses blockchain not only as a data repository but also as an execution platform for smart contracts that automate key processes in the credit evaluation workflow.

Key Results

The experimental results demonstrate the efficacy of the proposed system:

  • LSTM Performance: The LSTM model achieved superior performance in analyzing credit evaluation text compared to traditional machine learning methods like Support Vector Machine (SVM) and Naive Bayes (NB).
  • Accuracy and F1-Score: The system achieved high accuracy and F1-scores, consistently increasing with the scale of the corpus, reaching approximately 90% with sufficient data.
  • Model Comparison: In a direct comparison on the Chinese text dataset, the LSTM model significantly outperformed SVM and NB in sentiment analysis.
  • System Feasibility: The experiments confirmed that the blockchain-based credit evaluation system is feasible and efficacious for the food supply chain.

Limitations and Future Work

The authors acknowledge several limitations and suggest future research directions:

  • Binary Classification: The current system classifies credit evaluations into only two levels: "positive" and "negative." Future work should explore more exact and effective multi-class classification for a finer-grained evaluation.
  • Dataset Size and Diversity: The Chinese text dataset used was considered insufficient for comprehensive representation. Future research should train models on larger and more diverse datasets, including English datasets, to improve generalization.
  • Blockchain Scalability: The paper notes that blockchain's growing size and the need for each node to store the entire history remain significant challenges. Exploring solutions like double-chain architectures, as proposed by Leng et al., is suggested.
  • Further Exploration of Blockchain: The authors suggest that blockchain technology, being relatively new, still has a long way to go in terms of optimization and widespread adoption.

Summary updated

Abstract

The food supply chain is a complex system that involves a multitude of "stakeholders" such as farmers, production factories, distributors, retailers and consumers. "Information asymmetry" between stakeholders is one of the major factors that lead to food fraud. Some current researches have shown that applying blockchain can help ensure food safety. However, they tend to study the traceability of food but not its supervision. This paper provides a blockchain-based credit evaluation system to strengthen the effectiveness of supervision and management in the food supply chain. The system gathers credit evaluation text from traders by smart contracts on the blockchain. Then the gathered text is analyzed directly by a deep learning network named Long Short Term Memory (LSTM). Finally traders' credit results are used as a reference for the supervision and management of regulators. By applying blockchain, traders can be held accountable for their actions in the process of transaction and credit evaluation. Regulators can gather more reliable, authentic and sufficient information about traders. The results of experiments show that adopting LSTM results in better performance than traditional machine learning methods such as Support Vector Machine (SVM) and Navie Bayes (NB) to analyze the credit evaluation text. The system provides a friendly interface for the convenience of users.
BibTeX
@article{mao2018credit,
  author = {Dianhui Mao and Fan Wang and Zhihao Hao and Haisheng Li},
  title = {{Credit Evaluation System Based on Blockchain for Multiple Stakeholders in the Food Supply Chain}},
  journal = {International Journal of Environmental Research and Public Health},
  year = {2018},
  doi = {10.3390/ijerph15081627},
}
Figure