黃郁珊

Product quality is the focus of consumers' attention, and maintaining product quality is an important responsibility of the industry. The maintenance of food quality has received great attention from all walks of life because it involves the safety of human life. In order to maintain the food quality and safety of each company in the supply chain, companies must carefully select their raw material suppliers and carry out "self-management" within the company itself. Traditionally, the internal quality management analysis used by enterprises mostly uses Six Sigma Methodology, control charts, and fishbone diagrams to maintain quality and improve problems. However, these methods cannot detect quality abnormalities in real-time and react quickly and are prone to unsatisfactory results due to incomplete considerations.

With the emergence of the concept of the Internet of Things and the popularization of technology, many scholars use the collected production data and then use artificial intelligence and machine learning techniques to analyze the quality of production. However, the existing research is mostly limited to a single function, lacks a model that integrates the overall quality monitoring and management functions, and does not mention the method of continuous updating of the model. Data science provides a systematic cycle of mastering the current situation, predicting and preventing, diagnosing prescriptions, and continuously optimizing independently. Each supply point of the food supply chain can be regarded as a system, therefore, the concepts of system engineering "system monitoring" and "continuous improvement" and risk management "prevention" can be combined with the systematic analysis cycle of data science to design a "food safety monitoring and management model" with current status, monitoring variables, abnormal detection, trend prediction, early warning, diagnosis prescription, experience accumulation, and continuous optimization cycle to maintain the food quality and safety of all companies in the supply chain.

In response to the maintenance needs of food quality and safety, this research proposes a "food quality monitoring model with a continuous improvement mechanism" and develops its implementation technology. Aiming at this goal, this research designs a "food safety maintenance model" to analyze the functions that the food supply chain and the enterprise should have separately and jointly. Then, referring to the concept of data science, design a "food quality monitoring model with a continuous improvement mechanism", and based on this model, design a "data model for food quality monitoring and analysis" in response to the needs of food quality analysis. Finally, the use of statistics and machine learning techniques to achieve "food quality monitoring, prediction, and early warning functions" and "continuous improvement mechanism".

This study uses public data sets to conduct experiments to verify the feasibility and effectiveness of the proposed technology. In the method of anomaly detection, the accuracy rate of using the full data in a continuous improvement mechanism is 95.33%, and the accuracy of the method using the full data directly is 95.18%; the accuracy of using the sliding window method to obtain data and using it with a continuous improvement mechanism is 95.26%, and the accuracy of directly using the sliding window method to obtain data is 95.06%. In the trend prediction method, the root-mean-square error of using all data with a continuous improvement mechanism is 6.9, and the root-mean-square error of using all data directly is 9.8. In these two functions, the experimental results of the method with continuous improvement mechanism are better than the method of directly using incremental data, so the method with a continuous improvement mechanism is feasible and effective.