An ensemble technique for multi class imbalanced problem using probabilistic neural networks
- Authors: Chandrasekara, N. , Tilakaratne, Chandima , Mammadov, Musa
- Date: 2018
- Type: Text , Journal article
- Relation: Advances and Applications in Statistics Vol. 53, no. 6 (2018), p. 647-658
- Full Text: false
- Reviewed:
- Description: The class imbalanced problem is one of the major difficulties encountered by many researchers when using classification tools. Multi class problems are especially severe in this regard. The main objective of this study is to propose a suitable technique to handle multi class imbalanced problem. Probabilistic neural network (PNN) is used as the classification tool and the directional prediction of Australian, United States and Srilankan stock market indices is considered as the application. We propose an ensemble technique to handle multi class imbalanced problem that is called multi class undersampling based bagging (MCUB) technique. This is a new initiative that has not been considered in the literature to handle multi class imbalanced problem by employing PNN. The results obtained demonstrate that the proposed MCUB technique is capable of handling multi class imbalanced problem. Therefore, the PNN with the proposed ensemble technique can be used effectively in data classification. As a further study, other classification tools can be used to investigate the performance of the proposed MCUB technique in solving class imbalanced problems.
Modified neural network algorithms for predicting trading signals of stock market indices
- Authors: Tilakaratne, Chandima , Mammadov, Musa , Morris, Sidney
- Date: 2009
- Type: Text , Journal article
- Relation: Journal of Applied Mathematics and Decision Sciences Vol. 2009, no. (2009), p.
- Full Text: false
- Reviewed:
- Description: The aim of this paper is to present modified neural network algorithms to predict whether it is best to buy, hold, or sell shares (trading signals) of stock market indices. Most commonly used classification techniques are not successful in predicting trading signals when the distribution of the actual trading signals, among these three classes, is imbalanced. The modified network algorithms are based on the structure of feed forward neural networks and a modified Ordinary Least Squares (OLSs) error function. An adjustment relating to the contribution from the historical data used for training the networks and penalisation of incorrectly classified trading signals were accounted for, when modifying the OLS function. A global optimization algorithm was employed to train these networks. These algorithms were employed to predict the trading signals of the Australian All Ordinary Index. The algorithms with the modified error functions introduced by this study produced better predictions.