Caller behaviour classification: A comparison of SVM and FIS techniques

Accurate classification of caller interactions within Interactive Voice Response systems would assist corporations to determine caller behaviour within these telephony applications. This paper proposes a classification system with these capabilities. Fuzzy Inference Systems, Support Vector Machine and ensemble of field classifiers for a pay beneficiary application were developed. Accuracy, sensitivity and specificity performance metrics were computed and compared for these classification solutions. Ideally, a field classifier should have high sensitivity and high specificity. The Support Vector Machine field classifiers are the preferred models for the 'Say account', 'Select beneficiary' and 'Say confirmation' fields as these solutions yield the best performance results. However, the ensemble of field classifiers is the most accurate for the 'Say amount' field.