No abstract

Multilayer feed-forward artificial neural networks are one of the most frequently used data mining methods for classification, recognition, and prediction problems. The classification accuracy of a multilayer feed-forward artificial neural networks is proportional to training. A well-trained multilayer feed-forward artificial neural networks can predict the class value of an unseen sample correctly if provided with the optimum weights. Determining the optimum weights is a nonlinear continuous optimization problem that can be solved with metaheuristic algorithms. In this paper, we propose a novel multimean particle swarm optimization algorithm for multilayer feed-forward artificial neural networks training. The proposed multimean particle swarm optimization algorithm searches the solution space more efficiently with multiple swarms and finds better solutions than particle swarm optimization. To evaluate the performance of the proposed multimean particle swarm optimization algorithm, experiments are conducted on ten benchmark datasets from the UCI repository and the obtained results are compared to the results of particle swarm optimization and other previous research in the literature. The analysis of the results demonstrated that the proposed multimean particle swarm optimization algorithm performed well and it can be adopted as a novel algorithm for multilayer feedforward artificial neural networks training.

The interest in automatic gender classification has increased rapidly, especially with the growth of online social networking platforms, social media applications, and commercial applications. Most of the images shared on these platforms are taken by mobile phone with different expressions, different angles and low resolution. In recent years, convolutional neural networks have become the most powerful method for image classification. Many researchers have shown that convolutional neural networks can achieve better performance by modifying different network layers of network architecture. Moreover, the selection of the appropriate activation function of neurons, optimizer and the loss function directly affects the performance of the convolutional neural networks. In this study, we propose a gender classification system from facial images taken by mobile phone using convolutional neural networks. The proposed convolutional neural networks have a simple network architecture with appropriate parameters can be used when rapid training is needed with the amount of limited training data. In the experimental study, the Adience benchmark dataset was used with 17492 different images with different gender and ages. The classification process was carried out by 10-fold cross validation. According the experimental results, the proposed convolutional neural networks predicted the gender of the images 98.87% correctly for training and 89.13% for testing.

No abstract

Most of the machine learning and data mining algorithms use discrete data for the classification process. But, most data in practice include continuous features. Therefore, a discretization pre-processing step is applied on these datasets before the classification. Discretization process converts continuous values to discrete values. In the literature, there are many methods used for discretization process. These methods are grouped as supervised and unsupervised methods according to whether a class information is used or not. In this paper, we used two unsupervised methods: Equal Width Interval (EW), Equal Frequency (EF) and one supervised method: Entropy Based (EB) discretization. In the experiments, a well-known 10 dataset from UCI (Machine Learning Repository) is used in order to compare the effect of the discretization methods on the classification. The results show that, Naive Bayes (NB), C4.5 and ID3 classification algorithms obtain higher accuracy with EB discretization method.

No abstract

No abstract

The performance of the bat algorithm has been improved by updating the velocity equation.  The proposed weight-based bat algorithm (WBBA) hybrited with the k-mean (KM) clustering algorithm to enhance the performance of the clusting analyze.  The proposed WBBA-KM clustering method has been applied on commonly used UCI datasets and has been compared with twelve clustering method in the literature.  The proposed WBBA-KM clustering method has been showed very promising results and can be used to cluster real world problems.