Abstract

The risk of poisoning in processed fish products due to increased histamine content can occur due to fish processing not meeting handling and hygiene standards. Therefore, producers of processed fish must have a traceability system to ensure the food products' safety. This research aims to design a real-time business intelligence architecture for product traceability and recall decision support. Real-time consumer response data is critical to support tactical decision-making regarding traceability. Real-time data introduction into the company's data warehouse can be done based on consumer response input through social media. The TF-IDF (Term Frequency-Inverse Document Frequence) data mining technique is used to process text data by giving weight to each keyword related to customer complaints about processed fish products. The k-Nearest Neighbor (KNN) data mining technique is used for sentiment analysis of tweets that classify text data and keywords into positive and negative classes to produce information supporting traceability system decisions. Data sources are comments, location tags, and multimedia from the company's social media group members. Based on the review of the proposed business intelligence architecture, data mining operationalization and dashboard visualization can support the achievement of KPIs from the traceability perspective. Companies can use the information generated to support work performance measured through KPIs. The speed of identifying the location of the cause of histamine increase, accuracy in identifying batch ID products, speed in publishing traceability results, and speed in recalling products from the market are some of the KPIs of this traceability system.


Key Words: traceability, business intelligence, fish processing