News-informed and collaboration-based forecasting support system (NEWSLAB-FSS) for optimum integration of quantitative and judgmental forecasts

Judgmental forecast is a category of forecasting technique that depends on subjective thinking of the forecasters, such as their intuition, knowledge, and experience. On the other hand, the quantitative forecast is a type of forecast which is generated by procedures not involving human's judgment, such as forecasts produced by statistical methods, machine learning, etc. The comparison between judgmental and quantitative forecasts reveals that both methods possess complementary strengths. Thus, it is a good idea to integrate both of these techniques together in dealing with forecasting task. The integration of quantitative and judgmental forecasting methods has been increasingly applied to give better performance in forecasting. Judgmental adjustment is one of the few instances. It has been gaining recognition among forecasting practitioners because of its quick and convenient way to perform forecast. However, few researches have criticized this approach for its disadvantages, i.e. bias and inconsistency, which originate from the weaknesses in human thinking. In this thesis a forecasting framework is proposed, called the NEWSLAB-FSS. It adopts the judgmental adjustment approach and provides contextual information to the forecasters in order to reduce the encounter with bias and inconsistency by establishing domain knowledge. Hence, it is expected to produce more accurate forecast. The proposed framework comprises five different modules that are designed to convey the contextual information to the forecasters, i.e. time series graphical display, quantitative forecast, news-based supportive information, user comment and similarity-based pattern search. Finally, the proposed NEWSLAB-FSS was applied to predict stock price data. These experiments were performed with the help of thirty people, who were asked to perform forecasting tasks. The experiments aimed to make comparison between the effectiveness of the proposed framework and the effectiveness of the currently common approach of forecasting, i.e. the judgmental adjustment approach. The results revealed that the proposed framework was more effective than the current approach, since the test subjects who utilized the NEWSLAB-FSS to make forecasts produced more accui rate forecasts. The study concludes that it is important to incorporate domain knowledge into forecasting process. The proposed framework has also succeeded to integrate the domain knowledge into the current trend of forecasting process, which resulted in a more accurate forecast.