Association rule mining technique for psychometric personality testing and behaviour prediction

At the heart of personality psychology lies one single fundamental motive and that is to be able to anticipate how an individual will think, behave and feel at any future instant. Quite unfortunately this field has not been very successful in achieving this. Though this field has given us great insights about the working of the mind, cognitive processes and emotions, it has failed to accomplish its central objective i.e., to predict human behaviour. We propose in this paper a novel technique of predicting human behaviour without the need of any abstraction about the mind or its internal workings. We propose the use of simple and straightforward statistics for this purpose. Applying simple association rule mining on behaviours of thousands of people, association rules having high confidence values can be identified. And based on these rules, strong conclusions can be made in anticipating the behaviour of an individual. An analytical study was conducted on answers provided by 1414 candidates to a 163-question personality survey. The survey was based on the famous questionnaire prepared by Raymond Cattell. This survey was chosen to first try and prove the ambiguity in the current psychological concepts. Following that simple association rule mining was applied on the data to obtain associations between variables. The strongest association obtained with 97.2% confidence was an inter-class association rather than an intra- class association as would be expected from traditional psychology point-of-view.