Violence Detection Approach based on Cloud Data and Neutrosophic Cognitive Maps

Abstract

Violence has remained a momentous problem since time immemorial. Various scientific studies are conducted in the recent past to identify the stimuli causing violent behavior among the masses and to achieve the target of cloud data protection. Given the inherent ambiguity or indeterminacy in human behaviour, this study in the area of violence detection appears to be effective, as it finds a variety of stimuli and character qualities that contribute to violent conduct among masses. This uncertainty of traits causing violence can easily be seen in surveillance data present over the cloud and also from the data collected using academic research. Therefore, for the purpose of identifying violent behavior we have considered the factors (data) from existing research and from data over clouds. The factors that lead to violent behavior and are identified by algorithms running over clouds are termed as determinate or certain factors. The factors that were not considered and least identified by the cloud algorithms and given less importance are termed indeterminate factors or uncertain factors. The indeterminate factors are also considered based on the expert’s opinion where the experts are not in a condition to provide a clear stance or when they are neutral in their opinion. Tests are performed using Neutrosophic Cognitive Maps (NCMs) to model the violent behavior taking into consideration both determinate and indeterminate factors. Earlier these tests were performed using Fuzzy Cognitive Maps (FCMs) where indeterminate or uncertain factors were not considered. Therefore, we provide a brief comparison between NCMs and FCMs and show how effective NCMs are when we need to consider the uncertainty of concepts while carrying out tests for identifying violent behavior. Later results are obtained by forming a Neutrosophic adjacency matrix which is evaluated using the concepts of linear algebra. The obtained results in the form of 1 ∗ n vector (1 I I I I 1 I 1 I I I I I I I I I I I) clearly shows the presence of indeterminate factor ‘I’ in the vector which was absent in earlier models when designed using FCMs. This shows how these indeterminate or uncertain factors play a significant role in cultivating violent behavior which was not shown in the previous study. The study is significant since it takes into account factors from cloud data, experts’ opinions, and also from literature, and shows how these factors are taken into consideration at the data level itself so that they will not impact the modeling stage, and machine learning algorithms will perform well because uncertain and indeterminate information is taken care of at training phase itself. Hence uncertainty could be reduced in machine learning algorithms and in the overall recognition of violent behavior.