Heri Kuswanto

A new simple test against spurious long memory using temporal aggregation

We have developed a new test against spurious long memory based on the invariance of long memory parameter to aggregation. By using the local Whittle estimator, the statistic takes the supremum among combinations of paired aggregated series. Simulations show that the test performs good in finite sample sizes, and is able to distinguish long memory from spurious processes with excellent power. Moreover, the empirical application gives further evidence that the observed long memory in German stock returns is spurious.

Ensemble method based on ARIMA-FFNN for climate forecasting

Ensemble forecasting is one of relatively new modern methods for time series forecasting that employs averaging or stacking from the results of several methods. This paper focuses on the development of ensemble ARIMA-FFNN for climate forecasting by using averaging method. Two data about monthly rainfall in Indonesia, i.e. Wagir and Pujon region, are used as case study. Root mean of squares errors in training and testing datasets are used for evaluating the forecast accuracy. The results of ensemble ARIMA-FFNN are compared to one classical statistical method, i.e. individual ARIMA, and two modern statistical methods, namely individual FFNN and ensemble FFNN. The results show that ARIMA yields more accurate forecast in training datasets than other methods, whereas in testing datasets show that FFNN is the best method. Additionally, this conclusion in line with the results of M3 competition, i.e. modern methods or complex methods do not necessarily produce more accurate forecast than simpler one.

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.

Bayesian mixture model averaging for identifying the different gene expressions of chickpea (Cicer arietinum) plant tissue

ABSTRACT Identification of different gene expressions of chickpea (Cicer arietinum) plant tissue is needed in order to develop new varieties of chickpea plant which is resistant to disease through the insertion of genes. This plant is the third legume plant of the Leguminosae (Fabaceae) family and is much needed in the world due to its high-protein seeds and roots that contain symbiotic nitrogen-fixing bacteria. This paper has succeeded to demonstrate the work of Bayesian mixture model averaging (BMMA) approach to identify the different gene expressions of chickpea plant tissue in Indonesia. The results show that the best BMMA normal models contain from 727 (73%) up to 939 (94%) models from 1,000 generated mixture normal models. The fitted BMMA models to gene expression differences data on average is 0.2878511 for Kolmogorov–Smirnov (KS) and 0.1278080 for continuous rank probability score (CRPS). Based on these BMMA models, there are three groups of gene IDs: downregulated, regulated, and upregulated. The results of this grouping can be useful to find new varieties of chickpea plants that are more resistant to disease. The BMMA normal models coupled with Occams window as a data-driven modeling have succeed to demonstrate the work of building the gene expression differences microarray experiments 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.

Enzyme Classification on DUD-E Database Using Logistic Regression Ensemble (Lorens)

Discovery of drugs has been a complex process, time-consuming and expensive until an alternative of making drug has been found i.e. using in silico method to discover potential inhibitor. During the process of drug design, compound classification is carried out through docking score steps. The aim of this research is to predict the docking score results using proper methods for classification i.e. a computationally based method and a standard statistical method. This research examined three target enzymes listed in DUD-E database i.e. aofb, cah2 and hs90a. Each enzyme consists of different compounds that will be classified as good inhibitor (ligand) and bad inhibitor (decoy). In this research, the docking score step is conducted by binary logistic regression and logistic regression ensemble (Lorens). Binary logistic regression yields on 90.4% of accuracy for aofb, 91.7% for cah2 and 94% for hs90a enzyme. Meanwhile, logistic regression ensemble (Lorens) results on the accuracy levels of 88.95, 92.1 and 100% for aofb, cah2 and hs90a consecutively. This paper showed that logistic regression ensemble method outperforms standard logistic regression to be used for the inhibitor classification.

Multi Model Calibration of Rainfall Forecasts in East Nusa Tenggara Using Ensemble Model Output Statistics

The current weather changes uncertainly, marked by significant rise in surface temperatures and reduced rainfall in the tropics. The impact of this uncertainty often leads to a misprediction which may causes lack of anticipation for the upcoming extreme weather events. Statistical approach is required to reduce the error prediction. In 2015, Indonesia experienced the drought-related threat induced by the impact of El Nino storms in the Asia Pacific region. It affected the agricultural sector where almost 21 thousand hectares of agricultural land along Java, Bali and Nusa Tenggara were experiencing drought. This research calibrates ensemble forecasts to take into account the uncertainty and reduce the bias. The calibration of ensemble forecast is carried out by Ensemble Model Output Statistics (EMOS), which is applied to rainfall forecast in East Nusa Tenggara. The results show that calibration using EMOS is capable to produce a reliable forecats, in which the optimum forecast is obtained by training window of 24 months.

An Approach to Measure Financial Risk Relative Indices: A Case Study of Indonesian Insurance Companies

The number of insurance companies in Indonesia has been increasing over the years. This is a good indicator of the society’s awareness ofthe importance of insurance. In order to ensure the sustainability of acompany as well as its customers’ safety, the insurance companies have to maintain the quality of their performance. To deal with this, the Financial Service Authority (OJK) of Indonesia needs to monitor the risk level of each insurance company. For the company, financial risk analysis is required to formulate strategies for reducing its risk. Financial risk analysis is also important for prospective investors or creditors as one of the considerations to formulate a business plan with the company. This chapter introduces an approach to measure financial risk through financial performance data as reported in the company’s annual report. In this case, the risk variables in the balance sheet are available only on a yearly basis, and hence, a time series based approach for measuring risk, such as Value at Risk (VaR), cannot be applied. The limited number of series will explode the variance estimate of the parameter distribution used to calculate VaR. As an alternative, the risk can be measured by an index showing the financial risk of a company relative to the others. The fact that financial risk is a latent variable, which can be measured only through its indicators, leads to the idea of calculating the index by using the concept of the Confirmatory Factor Analysis (CFA). This chapter applies that idea to calculate the financial risk relative indices of life insurance companies in Indonesia. This approach offers another benefit, the ability to investigate variables which significantly contributes to increase the risk. The analysis shows that a company is said to have a high financial risk relative to the others if the index is below 0.29.

Population Analysis of Disabled Children by Departments in France

In this study, a statistical analysis is performed by model the variations of the disabled about 0-19 years old population among French departments. The aim is to classify the departments according to their profile determinants (socioeconomic and behavioural profiles). The analysis is focused on two types of methods: principal component analysis (PCA) and multiple correspondences factorial analysis (MCA) to review which one is the best methods for interpretation of the correlation between the determinants of disability (independent variable). The hierarchical cluster analysis (HCA) can be used to classify the departments according to their profile determinants. Analysis of variance or ANOVA is performed to know difference the between cluster and within cluster variances of two proxy data (AEEH and EN3-EN12). The PCA reduces 14 determinants of disability to 4 axes, keeps 80% of total information, and classifies them into 7 clusters. The MCA reduces the determinants to 3 axes, retains only 30% of information, and classifies them into 4 clusters. The ANOVA of the proxy data by department cluster are difference significant between cluster and the variance within of cluster is not difference significant, the cluster are homogeneous.

Bayesian mixture modeling for blood sugar levels of diabetes mellitus patients (case study in RSUD Saiful Anwar Malang Indonesia)

Bayesian statistics proposes an approach that is very flexible in the number of samples and distribution of data. Bayesian Mixture Model (BMM) is a Bayesian approach for multimodal models. Diabetes Mellitus (DM) is more commonly known in the Indonesian community as sweet pee. This disease is one type of chronic non-communicable diseases but it is very dangerous to humans because of the effects of other diseases complications caused. WHO reports in 2013 showed DM disease was ranked 6th in the world as the leading causes of human death. In Indonesia, DM disease continues to increase over time. These research would be studied patterns and would be built the BMM models of the DM data through simulation studies where the simulation data built on cases of blood sugar levels of DM patients in RSUD Saiful Anwar Malang. The results have been successfully demonstrated pattern of distribution of the DM data which has a normal mixture distribution. The BMM models have succeed to accommodate the real condition of the DM data based on the data driven concept.