Document representation based on probabilistic word clustering in customer-voice classification

Pattern Analysis and Applications - Customer-voice data have an important role in different fields including marketing, product planning, and quality assurance. However, owing to the manual...     

Smartphone user segmentation based on app usage sequence with neural networks

The term user segmentation refers to classifying users into groups depending on their specific needs, characteristics, or behaviors. It is a key element of product development and marketing in many industries, such as the smartphone industry, which employs user segmentation to gather information about usage logs, to produce new products for such specific groups of users. However, previous studies on smartphone user segmentation have been primarily based on demographics and reported usage, which are inherently subjective and prone to skew by the observers and participants. Hamka et al. (2014) was the first to conduct a study, in which smartphone user segmentation was performed using log data collected through smartphone measurements. However, they focused only on network usage and the number of apps used, and not on characteristics or preferences. In this study, we proposed novel ways of segmenting smartphone users based on app usage sequences collected from smartphone logs. We proposed a variant of seq2seq architecture combining the advantages of previous deep neural networks: neural embedding architecture and seq2seq architecture. Furthermore, we compared the user segmentation results of the proposed method with an answer set of segmentation results conducted by domain experts. These experiments demonstrated that the proposed method effectively determines similarities between usage sequences and outperforms existing user segmentation methods.     

A virtual metrology system for semiconductor manufacturing

Nowadays, the semiconductor manufacturing becomes very complex, consisting of hundreds of individual processes. If a faulty wafer is produced in an early stage but detected at the last moment, unnecessary resource consumption is unavoidable. Measuring every wafer’s quality after each process can save resources, but it is unrealistic and impractical because additional measuring processes put in between each pair of contiguous processes significantly increase the total production time. Metrology, as is employed for product quality monitoring tool today, covers only a small fraction of sampled wafers. Virtual metrology (VM), on the other hand, enables to predict every wafer’s metrology measurements based on production equipment data and preceding metrology results. A well established VM system, therefore, can help improve product quality and reduce production cost and cycle time. In this paper, we develop a VM system for an etching process in semiconductor manufacturing based on various data mining techniques. The experimental results show that our VM system can not only predict the metrology measurement accurately, but also detect possible faulty wafers with a reasonable confidence.     

Keystroke dynamics-based authentication for mobile devices

Recently, mobile devices are used in financial applications such as banking and stock trading. However, unlike desktops and notebook computers, a 4-digit personal identification number (PIN) is often adopted as the only security mechanism for mobile devices. Because of their limited length, PINs are vulnerable to shoulder surfing and systematic trial-and-error attacks. This paper reports the effectiveness of user authentication using keystroke dynamics-based authentication (KDA) on mobile devices. We found that a KDA system can be effective for mobile devices in terms of authentication accuracy. Use of artificial rhythms leads to even better authentication performance.     

Pattern selection for support vector regression based response modeling

Two-stage response modeling, identifying respondents and then ranking them according to their expected profit, was proposed in order to increase the profit of direct marketing. For the second stage of two-stage response modeling, support vector regression (SVR) has been successfully employed due to its great generalization performances. However, the training complexities of SVR have made it difficult to apply to response modeling based on the large amount of data. In this paper, we propose a pattern selection method called Expected Margin based Pattern Selection (EMPS) to reduce the training complexities of SVR for use as a response modeling dataset with high dimensionality and high nonlinearity. EMPS estimates the expected margin for all training patterns and selects patterns which are likely to become support vectors. The experimental results involving 20 benchmark datasets and one real-world marketing dataset showed that EMPS improved SVR efficiency for response modeling.     

Constructing belief networks from realistic data

In this paper, we report our investigation on the behavior of some Bayesian belief network learning methods, when applied to realistic case data that do not accurately reflect the probability distribution of the underlying domain. The investigation focuses on two types of such data of practical interest, namely data excluding null or normal cases and data contaminated by noise, and their effects on the performance of two learning methods: the K2 algorithm, a representative method of Bayesian approach, and the extended Hebbian learning (EHL) method, representing those of neural network approach. Both analytical and experimental results show that when null cases are removed from the case database, the EHL method is able to accurately construct the underlying causal structure; whereas K2 algorithm may fail to properly handle root nodes. On the other hand, the EHL method is noise sensitive while K2 has certain inherent noise-resistance capability. Suggestions to extend both K2 and EHL to better cope with these problems are also made, and their effectiveness tested through a systematic computer simulation experiment. ©1999 John Wiley & Sons, Inc.     

Smoothed Bagging with Kernel Bandwidth Selectors

Recently, a combined approach of bagging (bootstrap aggregating) and noise addition was proposed and shown to result in a significantly improved generalization performance. But, the level of noise introduced, a crucial factor, was determined by trial and error. The procedure is not only ad hoc but also time consuming since bagging involves training a committee of networks. Here we propose a principled procedure of computing the level of noise, which is also computationally less expensive. The idea comes from kernel density estimation (KDE), a non-parametric probability density estimation method where appropriate kernel functions such as Gaussian are imposed on data. The kernel bandwidth selector is a numerical method for finding the width of a kernel function (called bandwidth). The computed bandwidth can be used as the variance of added noise. The proposed approach makes the trial and error procedure unnecessary, and thus provides a much faster way of finding an appropriate level of noise. In addition, experimental results show that the proposed approach results in an improved performance over bagging, particularly for noisy data.     

Effects of varying parameters on properties of self-organizing feature maps

The behavior of self-organizing feature maps is critically dependent on parameters such as lateral connection radius, lateral inhibition intensity, and network size. With no theoretical guidelines for the choice of these parameters, they are usually selected through a trial-and-error process. In order to provide heuristic guidelines for future model designers as well as to give insight into which model features are responsible for specific aspects of maps, we systematically varied these parameters and studied their effects on the properties of a self-organizing feature map. The connectivity radius was found to determine the size of activation clusters quadratically. As the intensity of lateral inhibition was varied, feature patterns varied from stripe-like to clusters in the map, with other intermediate patterns also occurring. The number of clusters of each feature increased nonlinearly as the network size increased.     

Evaluating the reliability level of virtual metrology results for flexible process control: a novelty detection-based approach

The purpose of virtual metrology (VM) in semiconductor manufacturing is to support process monitoring and quality control by predicting the metrological values of every wafer without an actual metrology process, based on process sensor data collected during the operation. Most VM-based quality control schemes assume that the VM predictions are always accurate, which in fact may not be true due to some unexpected variations that can occur during the process. In this paper, therefore, we propose a means of evaluating the reliability level of VM prediction results based on novelty detection techniques, which would allow flexible utilization of the VM results. Our models generate a high-reliability score for a wafer’s VM prediction only when its process sensor values are found to be consistent with those of the majority of wafers that are used in model building; otherwise, a low-reliability score is returned. Thus, process engineers can selectively utilize VM results based on their reliability level. Experimental results show that our reliability generation models are effective; the VM results for wafers with a high level of reliability were found to be much more accurate than those with a low level.