Hybridizing Cuckoo Search with Levenberg-Marquardt Algorithms in Optimization and Training of ANNs for Mass Appraisal of Properties

摘要

Various algorithms, including particle swarm optimization (PSO), genetic algorithm (GA), ant colony algorithm (AC), cuckoo search (CS) algorithm, and firefly algorithm (FA) have been introduced to help optimize artificial neural networks (ANNs), speed up convergence and iteration rates, and escape from trapping into local optimum. However, despite the capabilities of these algorithms, it is only GA that has been utilized in the mass appraisal of properties. Therefore, in order to deal with problems of inconsistencies in appraisal/valuation estimates that sometimes occur during predictions, CS, a meta-heuristic algorithm is introduced into the mass appraisal industry. The proposed algorithm is combined with Levenberg-Marquardt (LM) and back propagation (BP) algorithms to test their effectiveness in the prediction of property values. We analyzed a dataset of 3,494 property transactions from the city of Cape Town, South Africa. The results indicate that CSLM and CSBP outperformed standalone the conventional BP algorithm in optimizing and training of ANN for mass appraisal of properties. This is reflected in the minimal error matrices predicted by both CSLM and CSBP algorithms.