Prevention of Unwanted Calls Over Telephony Network

Abstract

The main motive is to increase the malicious call prevention while not relying on any underlying architecture of the phone . Challenges faced here are -how to gather useful information on how to decrease benign calls from being blocked . The first part of the work is to gather data concerning nuisance callers so as to create an efficient interference mechanism, with the help of Machine Learning Algorithms. Gathering of the dataset is based on client statistics. In that, there are three varieties of lessons like telemarketing, Unwanted (Malicious), robocalls.If a call is Malicious or not is judged based on few functions .So compare the effectiveness of these capabilities with the aid of numerous cutting-edge models. For this, usage of both Area under curve (AUC) score, and layout a brand new metric, average first prediction (AFP) is done. Average first prediction is designed to assess the averaged amount of nuisance calls that desires to be located before a user can expect it as a nuisance caller, without affecting benign traffic. Evaluation shows that segregating the planned options, a random forest version can do Associate in Nursing AUC score of at 0.98 ; additionally , it reduces the averaged important determined nuisance calls via up to 88% from a black-listing method, whilst making sure that over 98% of the benign calls will now no longer be banned from connecting . The system would be fast, effective, efficient, UI would be lightweight ,and this would work without accessing the telephony network architecture.