Dedy Dwi Prastyo

Two levels ARIMAX and regression models for forecasting time series data with calendar variation effects

The aim of this research is to develop a calendar variation model for forecasting retail sales data with the Eid ul-Fitr effect. The proposed model is based on two methods, namely two levels ARIMAX and regression methods. Two levels ARIMAX and regression models are built by using ARIMAX for the first level and regression for the second level. Monthly men’s jeans and women’s trousers sales in a retail company for the period January 2002 to September 2009 are used as case study. In general, two levels of calendar variation model yields two models, namely the first model to reconstruct the sales pattern that already occurred, and the second model to forecast the effect of increasing sales due to Eid ul-Fitr that affected sales at the same and the previous months. The results show that the proposed two level calendar variation model based on ARIMAX and regression methods yields better forecast compared to the seasonal ARIMA model and Neural Networks.

T 2 Control Chart based on Successive Difference Covariance Matrix for Intrusion Detection System

The Intrusion detection is a process to monitor the events taking place in a computer system or network and analyze the monitoring results to find signs of intrusion. One of alternative solutions for intrusion detection is the usage of statistical methods that Statistical Process Control especially the control charts.. In this research, the Hotellings T 2 chart performance for intrusion detection is improved using the Successive Difference Covariance Matrix where the control limits will be calculated using Kernel Density Estimation. The proposed method using T 2 based on Kernel Density Estimation control limit outperforms other approaches both in training and testing dataset.

Multioutput Least Square SVR Based Multivariate EWMA Control Chart

Multioutput least square SVR has ability to remove serial correlation of process by mapping multivariate input space to multivariate output space. The aim of this research is to propose multioutput least squares SVR based multivariate EWMA control chart to monitor small change of multivariate autocorrelated process. VARMA model with additive and innovative outliers are generated to investigate the performance of proposed control chart. Simulation studies empirically show that multioutput least squares SVR based multivariate EWMA control chart detect either single or consecutive additive outlier takes place at different time in each variable accurately. On the contrary, single innovative outlier in each variable that occurs either at different time or at the same time is detected by multioutput least squares SVR based multivariate EWMA control chart as double out-of control signals.

T2 hotelling fuzzy and W2 control chart with application to wheat flour production process

A good product must satisfy the standard of corresponding quality characteristic. A product is categorized as defect when its quality characteristic out of its specification limit. The measurement of quality characteristic may produce error because it measures quality around its actual value instead of the actual value itself. This measurement error can produce an ambiguity. Some approaches have been developed to tackle this ambiguity problem. This research employed T2 Hotelling Fuzzy and W2 control chart with probability approach to deal with this issue. The proposed methods were applied to monitor the wheat flour production process. Three variables, multivariate processes, were used to identify the product quality such as moisture, gluten, and ash. The membership function used fuzzy approach is a combination of trapezoidal and linear curves. This research concluded that W2 control chart with probability approach performed better than T2 Hotelling Fuzzy control chart.

Bootstrap-based maximum multivariate CUSUM control chart

ABSTRACT Maximum multivariate cumulative sum (Max-MCUSUM) is one of the single control charts that plot single statistic as a representation of mean vector and covariance matrix. The Max-MCUSUM statistic has unknown specific distribution. The objective of this paper is to propose bootstrap-based Max-MCUSUM control chart for which reference value is predetermined in Phase I monitoring process. For various numbers of quality characteristics and correlation coefficients, the control limits estimated using bootstrap approach are presented in this paper. Furthermore, the average run lengths of bootstrap-based Max-MCUSUM control chart prove that the proposed control chart tends to be effective for monitoring the small shift in both mean and variance of a process. The illustrative examples are provided to demonstrate the applications of the proposed control chart for both simulation and real data.

Multivariate control chart based on PCA mix for variable and attribute quality characteristics

ABSTRACT Two types of control charts exist based on different quality characteristics: variable and attribute. These characteristics are commonly monitored using separate procedures. Only a few studies focused on the utilization of control charts to monitor a process with mixed characteristics. This study develops a new concept of the control chart based on a Principal Component Analysis (PCA) Mix, that is a PCA method that can jointly handle continuous and categorical data. The Kernel Density Estimation (KDE) method is used to estimate the control limit. Through simulation studies, the performance of the proposed chart is evaluated using the Average Run Length (ARL). control limits obtained from KDE produce a stable ARL0 at ~ 370 for For the shifted process, the proposed chart demonstrates excellent performance for an appropriate number of principal components used. Applications of the simulated process and real cases show that the proposed chart is sensitive to monitoring the shifted process.

Hybrid ARIMAX Quantile Regression Model for Forecasting Inflow and Outflow of East Java Province

Most of inflow and outflow data in Indonesia are characterized by trend, seasonal, calendar variation, and heterogeneous variance. This study proposed hybrid ARIMAX Quantile Regression model for forecasting data that have trend, seasonal, calendar variation, and heterogeneous variance. There are two types of data that we used in this research, i.e. simulation and real data. The real data are monthly inflow and outflow of Bank Indonesia at East Java Province per currency for the period 2003 to December 2016. There are three types of ARIMAX Quantile Regression models with different predictors that be used for forecasting both data. The results show that hybrid ARIMAX Quantile Regression model can capture accurately all patterns in the data. Moreover, this hybrid model yield better forecast than individual ARIMAX model at 8 of 14 currencies of inflow and outflow data in East Java Province. Thus, based on forecast accuracy criteria, i.e. RMSE, MAE and MdAE, it could be concluded that hybrid ARIMAX Quantile Regression tend to give better forecast than other individual method.

Multioutput least square SVR-based multivariate EWMA control chart: The performance evaluation and application

Abstract Autocorrelation leads to a bias estimator of standard control charts. It is important to develop control chart that allows autocorrelation and to evaluate its performance. The objective of this paper is to evaluate the performance of multioutput least square support vector regression (MLS-SVR)-based multivariate exponentially weighted moving average (MEWMA) control chart for monitoring multivariate autocorrelated data. For first order of vector autoregressive (VAR) and first order of vector moving average data, the proposed control chart tends to yield stable in-control average run length at about 200. The proposed control chart becomes more insensitive due to the increase of MEWMA smoothing parameter. In the real application, the proposed method is successfully applied to monitor water turbidity and chlorine residual data in the drinking water manufacturing process.

Model Selection in Feedforward Neural Networks for Forecasting Inflow and Outflow in Indonesia

The interest in study using neural networks models has increased as they are able to capture nonlinear pattern and have a great accuracy. This paper focuses on how to determine the best model in feedforward neural networks for forecasting inflow and outflow in Indonesia. In univariate forecasting, inputs that used in the neural networks model were the lagged observations and it can be selected based on the significant lags in PACF. Thus, there are many combinations in order to get the best inputs for neural networks model. The forecasting result of inflow shows that it is possible to testing data has more accurate results than training data. This finding shows that neural networks were able to forecast testing data as well as training data by using the appropriate inputs and neuron, especially for short term forecasting. Moreover, the forecasting result of outflow shows that testing data were lower accurate than training data.

Additive survival least square support vector machines: A simulation study and its application to cervical cancer prediction

Support Vector Machines (SVMs) has been widely applied for prediction in many fields. Recently, SVM is also developed for survival analysis. In this study, Additive Survival Least Square SVM (A-SURLSSVM) approach is used to analyze cervical cancer dataset and its performance is compared with the Cox model as a benchmark. The comparison is evaluated based on the prognostic index produced: concordance index (c-index), log rank, and hazard ratio. The higher prognostic index represents the better performance of the corresponding methods. This work also applied feature selection to choose important features using backward elimination technique based on the c-index criterion. The cervical cancer dataset consists of 172 patients. The empirical results show that nine out of the twelve features: age at marriage, age of first getting menstruation, age, parity, type of treatment, history of family planning, stadium, long-time of menstruation, and anemia status are selected as relevant features that affect the survival time of cervical cancer patients. In addition, the performance of the proposed method is evaluated through a simulation study with the different number of features and censoring percentages. Two out of three performance measures (c-index and hazard ratio) obtained from A-SURLSSVM consistently yield better results than the ones obtained from Cox model when it is applied on both simulated and cervical cancer data. Moreover, the simulation study showed that A-SURLSSVM performs better when the percentage of censoring data is small.Support Vector Machines (SVMs) has been widely applied for prediction in many fields. Recently, SVM is also developed for survival analysis. In this study, Additive Survival Least Square SVM (A-SURLSSVM) approach is used to analyze cervical cancer dataset and its performance is compared with the Cox model as a benchmark. The comparison is evaluated based on the prognostic index produced: concordance index (c-index), log rank, and hazard ratio. The higher prognostic index represents the better performance of the corresponding methods. This work also applied feature selection to choose important features using backward elimination technique based on the c-index criterion. The cervical cancer dataset consists of 172 patients. The empirical results show that nine out of the twelve features: age at marriage, age of first getting menstruation, age, parity, type of treatment, history of family planning, stadium, long-time of menstruation, and anemia status are selected as relevant features that affect the surviv...