Abstract—The paper proposes a unique procedure to reduce risk factors in risk assessment. It offers a variant of decision tree called Identification tree for reducing number of risk factors used in assessment. The model, which uses auto insurance as a case study, employs historical evidences of different vehicles as risk factors. The work offers reduction of original risk factors from a set of twenty three to a reduced set of nine risk factors. The model was validated using real time and industry specific data. Index Terms—Average disorder score, identification tree, risk assessment, risk factor. I. INTRODUCTION We come across many uncertainties in our daily lives. Some of these uncertainties might result in financial losses. Insurance, which safeguards these financial losses to a minimum, has many forms like life, home, auto, property etc to name a few [1]. RISK ASSEMENT ASSET THREAT VULNERABILITY MITIGATION What are you trying to protect? What are you afraid of happening? How could the threat occur? What is currently reducing the risk? Impact/Severity What is the impact to the business? 1. Negligible 2.Minor 3. Moderate 4. Major 5. Critical 6.Catastrophic Probability/Likelihood How likely is the threat? 1. Unforeseeable 2. Very unlikely 3. Possible 4. Likely 5. Very likely 6.Almost certain RISK LOG Priority hazard Occurrence of an event Impact (1-6) Impact of event on Probability (1-6) Probability of the event Risk Rating = (Impact*probability) Risk Rating = Impact * Probability Fig. 1. Risk assessment Risk assessment in insurance is a practice wherein vulnerability of an asset against a risk is measured [2]. This involves finding likelihood of a financial loss caused by Manuscript received October 25, 2012; revise March 2, 2013. Ankit Agarwal and Jyotsna Dongerdive are with the University Department of Computer Science, University of Mumbai, India (e-mail: ankit.g.agarwal@gmail.com, jyotss.d@gmail.com). Siby Abraham is with the Department of Mathematics & Statistics, Guru Nanak Khalsa College, University of Mumbai, India (e-mail: sibyam@gmail.com). various risk factors. The likelihood and impact of these risk factors are collected into a risk log for assessment. The various stages involved in risk assessment are given in Fig. 1 [2]. Conventionally, it is done using various statistical and computational techniques [3]. Almost all these techniques assume availability of all risk factors. The paper proposes a methodology to reduce risk factors. This uses an identification tree [4], [5] based approach to reduce all the risk factors to the most significant ones. The work, which uses auto insurance sector as a case study, proposes that risk assessment can be realized using this reduced set of risk factors. The paper is organized in six sections. Section II gives related work. Section III introduces the model proposed. Section IV provides the implementation. Section V gives experimental results and section VI offers conclusion and future work. II. RELATED WORKS There have been many attempts to discuss risk assessment computationally. Jianbing Xiahou and Yang Mu [6] used Decision Tree as a Data Mining technique to classify and select risk factors. They concluded that Data Mining approach gave better results than the conventionally used statistical technique called General Linear Model [7]. Yong Che et al. [8] offered an alternative model for risk assessment using Ubiquitous Computing. It was a three step process containing Clustering on input data using Adaptive Resonance Theory [9], a modification phase of the feature vectors and a Back Propagation Neural Network. Chin- sheng haung et al. [10] proposed an evaluation model for selecting insurance policy using Analytic Hierarchy Process [11] and Fuzzy Logic [12]. They used four variables to evaluate purchase of life insurance and annuity insurance including age, annual income, educational level and risk preference. Anna Jurek and Danuta Zakrzewska [13] used a Naive Bayes model along with Clustering technique for risk assessment in life insurance. The work involved classification of artificially generated data sets into three risk classes, which was further enhanced using Clustering. Arnold F. Shapiro [14] gave an overview of Soft Computing applications in Actuarial Science. The work covered important Soft Computing techniques like Neural Networks [15], Fuzzy Logic and Genetic Algorithms [16]. He [17] also provided a comprehensive overview of recent advances in the theory and implementation of intelligent and other computational techniques in insurance. Chin-Sheng-Huang [18] et al. used Decision Trees [19] to establish Decision models for five different insurance sectors in Taiwan. Almost all these methods assumed the existence of a set of exhaustive risk factors. Ankit Agarwal, Jyotshna Dongardive, and Siby Abraham Reduction of Risk Factors in Risk Assessment: An Identification Tree Approach International Journal of Modeling and Optimization, Vol. 3, No. 2, April 2013 167 DOI: 10.7763/IJMO.2013.V3.260