Abstract:
The current study tests and compares the performance of Feed Forward Neural Network (FFNN) and Recurrent Neural Network (RNN) using ten Artificial Neural Network (ANN) training algorithms in tenants' classification in the Nigerian residential rental market with Ikeja as case study. A total of 724 data samples of the prospective residential rental applicants was obtained from the databases of 53 property management firms in the Ikeja, Lagos metropolitan property market, Nigeria. A combined confusion matrix measures which include sensitivity, specificity, Type 1 and type 2 error, Mean Squared Error, Geometric mean, Precision, F-measure, F2 measure, Adjusted F-measure, Mathew's correlation coefficient, Area under the curve, training time, Epoch, and regression plots were used in evaluating the performance of both models and their respective training algorithms. The results of the descriptive analysis reveal that most property managers use the subjective method of tenant selection and are not aware of the use of decision Support systems in tenant selection. It was also found that Gender, Marital status, Family size, Education, Occupation, Income, References, Relationship with the property manager, and Tenant History significantly determine tenant selection. Findings of the NN analysis show among others that both FFNN and RNN achieved satisfactory performance, however, there is no significant difference in the performance of FFNN and RNN in terms of training time, Epoch and MSE, it was also discovered that Bayesian Regularization training algorithm outperformed other training algorithms in both models across all performance measures. The study concluded that BR- trained FFNN may most be suitable for tenant selection tasks in the Nigerian residential rental property markets.
Keywords; Decision support system, Tenant selection, Performance, Feed-forward, Recurrent, Training algorithms
