Improving residential housing project purchase by using integrated multi-attribute decision making and sentiment analysis technique

The residential house purchase decision making is highly complex due to reasons such as conflicting criteria which is hard to model, infrequent type of decisions, uncertain and irreversible decision outcomes, high investment, and long-term financial burden. Unlike many other types of purchasing, housing purchase decision-making is riskier and sometimes even ‘traumatic’. It is often associated with feeling of regret and the possibility of loss among homebuyers. Typically, the Multi Attribute Decision Making (MADM) models are used to systematically assist and structure residential housing project selection decision making. However, the MADM models impose deficiencies in the evaluation process due to insufficient knowledge of homebuyers, ignorance of public opinions and limited sources of information. Furthermore, the application of MADM models requires homebuyer to rely on their evaluation experience which potentially led to an imprecise decision. Hence, this study developed an improved model by integrating MADM and three approaches of Sentiment Analysis to capture and rank criteria from public opinions through online reviews. Properties online forums and google reviews were selected to extract public opinions through online reviews. Three high-rise residential projects located in Malaysia were used as case projects for demonstrating the model development and validation of the proposed framework. Three Sentiment Analysis approach were considered; Lexicon, Machine Learning and hybrid. Based on the ranking established by the models, it shows that location, facility, and house attributes are the most important criteria in residential housing purchase decision making. In addition, classification using a hybrid MADM Sentiment Analysis approach outperforms the Lexicon approach with better accuracy. The developed model can assist homebuyer in making decision for the current practice. Moreover, it can be generalised to other related multi-criteria applications with the use of online public opinions as reference.