Songyot Nakariyakul

This paper addresses the high-dimensional classification problem, which is very important in machine learning. When the number of features of the data is very high, the classification performance of a given classifier can degrade because there are not enough samples for training. One of the solutions to cope with this problem is to perform feature selection to reduce the number of features. We propose a new hybrid feature selection algorithm based on interaction information that improves upon the previous one. Our improved method employs interaction information to select candidate features to be added to the current feature subset. Cohen’s d is used as the significance testing to decide whether a new feature is permanently added to the subset. We adopt new stopping criteria to allow intensive search. Our search method is efficient and is able to find excellent solutions. Experiments results on eleven high-dimensional data sets show that compared to other hybrid feature selection algorithms, our proposed algorithm provides high classification accuracy and requires a small number of features for classification.

We address gene selection and machine learning methods for cancer classification using microarray gene expression data. Due to the high dimensionality of microarray data, traditional gene selection algorithms are filter-based, focusing on intrinsic properties of the data such as distance, dependency, and correlation. These methods are fast but select far too many genes to use for the classification task. In this work, we present a new hybrid filter-wrapper gene subset selection algorithm that is an improved modification of our prior algorithm. Our proposed method employs interaction information to rank candidate genes to add into a gene subset. It then conditionally adds one gene at a time into the current subset and verifies whether the resultant subset improves the classification performance significantly. Only significant genes are selected, and the candidate gene list is updated every time a gene is added to the subset. Thus, our gene selection algorithm is very dynamic. Experimental results on ten public cancer microarray data sets show that our method consistently outperforms prior gene selection algorithms in terms of classification accuracy, while requiring a small number of selected genes.

This paper introduces a new Android navigation app called NavTU for Thai people with visual impairments. Our NavTU app is designed to promote safe and independent travel for Thai people who are blind or have low vision. It is the integration of the GPS-based and camera-based technologies that not only helps the blind pedestrians navigate outdoors from place to place but also locates obstacles such as concrete electric poles, so that the visually impaired can avoid walking into them. The main advantage of NavTU is that the users are allowed to record their own walking paths to the preferred locations. This enables the blind users to walk through any local roads that might not be shown on Google Maps. The blind pedestrians can plan any outdoor path that is convenient and safe to them. The proposed app has Thai as the natural language interface, so it is easy for Thai people with visual impairments to use.

Many Thai people with visual impairments make a living by selling lottery tickets. To assist the blind in selling their tickets, we present a new Android app called LottoTU for scanning and reading aloud Thai lottery numbers for the blind. Our system includes lottery number segmentation, thresholding using Otsu's method, digit recognition using the Tesseract engine, and text-to-speech synthesis. Since lottery ticket images can be taken under different illumination conditions and at various orientations and distances to the smartphone camera, we implement three different sets of threshold values and morphological structuring element sizes to detect lottery number locations. The segmented lottery number images are thresholded using Otsu's method and converted into text data using the Tesseract OCR engine. The lottery number text data are then read aloud by text-to-speech software. The experimental results show that our LottoTU app successfully scans and reads aloud lottery numbers with high accuracy.

Gene selection is an important pre-processing step in microarray analysis and classification. While traditional gene selection algorithms focus on identifying relevant and irredundant genes, we present a new gene selection algorithm that chooses gene subsets based on their interaction information. Many individual genes may be irrelevant with the class, but when combined together, they can interact and provide information useful for classification. Our proposed gene selection algorithm is tested on four well-known cancer microarray datasets. Initial results show that our algorithm selects effective gene subsets and outperforms prior gene selection algorithm in terms of classification accuracy.