ABSTRACT
Canine personality and behavioural characteristics have a significant influence on relationships between domestic dogs and humans as well as determining the suitability of dogs for specific working roles. As a result, many researchers have attempted to develop reliable personality assessment tools for dogs. Most previous work has analysed dogs' behavioural patterns collected via questionnaires using traditional statistical analytic approaches. Artificial Intelligence has been widely and successfully used for predicting human personality types. However, similar approaches have not been applied to data on canine personality. In this research, machine learning techniques were applied to the classification of canine personality types using behavioural data derived from the C-BARQ project. As the dataset was not labelled, in the first step, an unsupervised learning approach was adopted and K-Means algorithm was used to perform clustering and labelling of the data. Five distinct categories of dogs emerged from the K-Means clustering analysis of behavioural data, corresponding to five different personality types. Feature importance analysis was then conducted to identify the relative importance of each behavioural variable's contribution to each cluster and descriptive labels were generated for each of the personality traits based on these associations. The five personality types identified in this paper were labelled: "Excitable/Hyperattached", "Anxious/Fearful", "Aloof/Predatory", "Reactive/Assertive", and "Calm/Agreeable". Four machine learning models including Support Vector Machine (SVM), K-Nearest Neighbour (KNN), Naïve Bayes, and Decision Tree were implemented to predict the personality traits of dogs based on the labelled data. The performance of the models was evaluated using fivefold cross validation method and the results demonstrated that the Decision Tree model provided the best performance with a substantial accuracy of 99%. The novel AI-based methodology in this research may be useful in the future to enhance the selection and training of dogs for specific working and non-working roles.
Subject(s)
Algorithms , Artificial Intelligence , Humans , Dogs , Animals , Bayes Theorem , Machine Learning , PersonalityABSTRACT
Neuro Linguistic Programming (NLP) is a methodology used for recognition of human behavioral patterns and modification of the behavior. A significant part of this process is influenced by the theory of representational systems which equates to the five main senses. The preferred representational system of an individual can explain a large part of exhibited behaviors and characteristics. There are different methods to recognize the representational systems, one of which is to investigate the sensory-based words in the used language during the conversation. However, there are difficulties during this process since there is not a single reference method used for identification of representational systems and existing ones are subject to human interpretations. Some human errors like lack of experience, personal judgment, different levels of skill and personal mistakes may also affect the accuracy and reliability of the existing methods. This research aims to apply a new approach that is to automate the identification process in order to remove human errors, thereby increasing the accuracy and precision. Natural Language Processing has been used for automating this process, and an intelligent software has been developed to identify the preferred representational system with increased accuracy and reliability. This software has been tested and compared to human identification of representational systems. The results of the software are similar to a NLP practitioner, and the software responds more accurately than a human practitioner in various parts of the process. This novel methodology will assist the NLP practitioners to obtain an improved understanding of their clients' behavioral patterns and the associated cognitive and emotional processes.
