Classification and prediction on school children for food intake attitude toward food and beverage advertising on television: KFC as a case study / Ahmad Fikri Anuar

Serious health problem in adulthood stage such as diabetes, hypertension, cardiovascular diseases are related to obesity in early childhood. Obesity has become a problem in Malaysia in context of healthy lifestyle and in estimation, Malaysia has highest rates of obesity in South-East Asia involving children. One of the most dominant mediums that promote unhealthy foods is through Television Food Advertising (TVFA) that aimed for children. A new approach were applied by using Artificial Intelligence (AI) strategy, from that the Naive Bayes (NB) technique is used to predict the eating behaviour of children toward TVFA. Agile methodology is used as the project framework of the project study. Phase in agile is proceed one by one for each 5 phase of Agile. First phase is the Planning Phase where problem are identified, then the Analysis Phase to gather information about project, then Development Phase to design the system and produce prototype, followed with Testing Phase where all testing is done and lastly is to compile project finding in final year report as in the Documentation Phase. Five independent variables used in the model, are advertisement recognition, favourite advertisement, purchase request, product prefers and time watched TV. About 105 of school children of SK Merlimau of age 12 years old have been chosen as the target subject to realize the objectives of the prediction model. 80% of data collected were used as training data, and 20% were for the new data to be tested. 31 prediction models were produced by using this technique, and the results indicate that 78% accuracy from the data learnt. Although the accuracy result is not as expected (80% and above ), Naive Bayes could be implemented and may be continued by using other methods such as Support Vector Machine and Artificial Neural Network. The result finding for the system functionality is at best and functioning well. System can predict the expected outcome as data is learned before with appropriate variable. In the near future, hopefully there will be an extended work in terms of different technique and independent variables used to increase the accuracy.