Noor Akhmad Setiawan

Missing Attribute Value Prediction Based on Artificial Neural Network and Rough Set Theory

In this research, artificial neural network (ANN) combined with rough set theory (RST), named as ANNRST, is proposed to predict missing values of attribute. The prediction of missing values of attribute is applied on heart disease data from UCI datasets. The ANN used is multilayer perceptron (MLP) with resilient back-propagation learning. RST can reduce the dimensionality of attributes through its reduct. Reduct is used as input of ANN combined with decision attribute. By simulating of missing values, the prediction accuracy of ANN is compared to ANNRST. The accuracy of ANNRST is also compared with missing data imputation ofk-Nearest Neighbor (k-NN), most common attribute value method and ANN with piecewise linear network-orthonormal least square feature selection (PLN-OLS). Simulation results show that ANNRST can predict the missing value with maximum accuracy close to ANN without dimensionality reduction (pure ANN) and outperform k-NN, most common attribute value method, and ANN with PLN-OLS.

Missing data estimation on heart disease using Artificial Neural Network and Rough Set Theory

The objective of this research is to implement a method for estimating the real missing data in heart disease datasets and to show how it affects the resulting knowledge. Missing data is common problem in knowledge discovery from database (KDD) processes that can lead significant error in extracted knowledge. We use hybridization of artificial neural network and rough set theory (ANNRST) to estimate the real missing data on heart disease from UCI (University of California, Irvine) datasets. ANN with reduced input features is used to estimate the missing data. RST is used to reduce the dimensionality of input features and to extract the knowledge as reducts and rules from heart disease datasets with estimated missing data. RST, decomposition tree, local transfer function classifier (LTF-C) and k-nearest neighbor (k-NN) classifier are used to calculate the accuracy. Comparative study with k-NN estimation, most common attribute value filling and deletion of missing data are made to evaluate the extracted knowledge. ANNRST can be considered as the appropriate estimation method when strong relationship between original complete datasets and estimated datasets is important (the estimated datasets really represent the nature of original complete datasets) as it gives the best accuracy and coverage for almost all the classifiers.

Fuzzy Decision Support System for Coronary Artery Disease Diagnosis Based on Rough Set Theory

The objective of this research is to develop an evidence based fuzzy decision support system for the diagnosis of coronary artery disease. The development of decision support system is implemented based on three processing stages: rule generation, rule selection and rule fuzzification. Rough Set Theory (RST) is used to generate the classification rules from training data set. The training data are obtained from University California Irvine (UCI) data repository. Rule selection is conducted by transforming the rules into a decision table based on unseen data set. Furthermore, RST attributes reduction is proposed and applied to select the most important rules. The selected rules are transformed into fuzzy rules based on discretization cuts of numerical input attributes and simple triangular and trapezoidal membership functions. Fuzzy rules weighing is also proposed and applied based on rules support on the training data. The system is validated using UCI heart disease data sets collected from the U.S., Switzerland and Hungary and data set from Ipoh Specialist Hospital Malaysia. The system is verified by three cardiologists. The results show that the system is able to give the approximate possibility of coronary artery blocking.

Power transformer incipient faults diagnosis using Dissolved Gas Analysis and Rough Set

Dissolved Gas Analysis (DGA) is standard technique to detect and diagnose power transformer incipient faults. Many methods based on DGA have been proposed such as Duval Triangle, IEC, Rogers Ratio, Key Gases, etc. The relationship between gas and type of faults is difficult to model and highly non-linear. Knowledge Discovery from Data (KDD) based on Rough Set Theory (RST) can be used to find that relationship. Thus RST is used in this research. The objective of this research is to diagnose incipient fault of power transformer. The diagnosis method is using DGA and RST. All of possible gas ratios are used as input to determine types of fault. The value of gas ratios are discretized before being processed with Rough Set Theory (RST). The number of input attributes is reduced using RST. The knowledge is extracted in the form of IF-THEN rules. The extracted and reduced rules are used to diagnose the incipient faults of power transformer. The resulting rules have the accuracy of 81.25%.

A Comparative Study of Imputation Methods to Predict Missing Attribute Values in Coronary Heart Disease Data Set

The objective of this research is to investigate the effects of missing attribute value imputation methods on the quality of extracted rules when rule filtering is applied. Three imputation methods: Artificial Neural Network with Rough Set Theory (ANNRST), k-Nearest Neighbor (k-NN) and Concept Most Common Attribute Value Filling (CMCF) are applied to University California Irvine (UCI) coronary heart disease data sets. Rough Set Theory (RST) method is used to generate the rules from the three imputed data sets. Support filtering is used to select the rules. Accuracy, coverage, sensitivity, specificity and Area Under Curve (AUC) of Receiver Operating Characteristics (ROC) analysis are used to evaluate the performance of the rules when they are applied to classify the complete testing data set. Evaluation results show that ANNRST is considered as the best method among k-NN and CMCF.

Real-time traffic classification with Twitter data mining

The growth of vehicles in Yogyakarta Province, Indonesia is not proportional to the growth of roads. This problem causes severe traffic jam in many main roads. Common traffic anomalies detection using surveillance camera requires manpower and costly, while traffic anomalies detection with crowdsourcing mobile applications are mostly owned by private. This research aims to develop a real-time traffic classification by harnessing the power of social network data, Twitter. In this study, Twitter data are processed to the stages of preprocessing, feature extraction, and tweet classification. This study compares classification performance of three machine learning algorithms, namely Naive Bayes (NB), Support Vector Machine (SVM), and Decision Tree (DT). Experimental results show that SVM algorithm produced the best performance among the other algorithms with 99.77% and 99.87% of classification accuracy in balanced and imbalanced data, respectively. This research implies that social network service may be used as an alternative source for traffic anomalies detection by providing information of traffic flow condition in real-time.

Benchmarking of feature selection techniques for coronary artery disease diagnosis

Coronary artery disease (CAD) is a disease that causes many deaths in human. CAD occurs when the atherosclerosis (fatty deposits) blocks blood flow to the heart muscle in the coronary arteries. The gold standard method to diagnose CAD is coronary angiography. However, this method is invasive, risky and costly. Therefore, it is necessary to develop a method for diagnosing the CAD before coronary angiography is performed. The objective of this research is to provide a benchmark comparison of the feature selection techniques in the diagnosis of CAD. A total of four feature selection methods are used. These methods are motivated feature selection (MFS), correlation based feature selection (CFS), wrapper based feature selection (WFS) and rough set based feature selection (RST). The Naïve Bayes and J48 classifiers are used to diagnose the presence of CAD. The result shows that WFS and CFS are superior compared to MFS and RST.

Dynamic uav path planning for moving target intercept in 3D

Dynamic path planning is one of the challenging research problems which is needed to guide UAV (Unmanned Aerial Vehicle) for moving target intercept. This research is to develop an algorithm for moving target intercept with obstacle avoidance in 3D. The algorithm which is called L+Dumo Algorithm integrate a modified Dubins Algorithm and Linear Algorithm. The simulation is conducted using one UAV and one moving target with an obstacle of cylindrical shape in between both objects. The result shows that the algorithm can guide UAV to approach the moving target with the accuracy of 67.417% L+Dumo algorithm will solve dynamic path planning problem for moving target intercept in 3D with minimum complexity of computation.

A comparison of classification methods on diagnosis of thyroid diseases

Thyroid gland is one of the endocrine glands in the human body which produces thyroid hormone. This gland actively produces two kinds of hormone, namely thyroxine (T4) and triiodothyronine (T3). These hormones aim to produce protein, govern body metabolism, as well as to control body temperature circulation. Either excess or lack of these hormones will disturb those activities. The condition of excessive hormones is called hyperthyroid while the condition of lacking hormones is called hypothyroid. The major factor that influences the volume of the produced T3 and T4 hormones is iodine, because it is the main building-block substance of those hormones. The imbalance condition of this substance prevents thyroid to work properly. To identify the type of thyroid (normal, hypothyroid, hyperthyroid), WEKA (Waikato Environment for Knowledge Analysis) machine learning software is utilized. The thyroid dataset is taken from UCI (University of California - Irvine) machine learning repository as many as 215 instances. The test result shows that among six different methods available in WEKA, MLP (Multilayer Perceptron) method gives result with the highest accuracy, up to 96.74%, while BPA (Back Propagation Algorithm) methods produces result with the lowest accuracy, of 69.77%.

Automated segmentation of optic disc area using mathematical morphology and active contour

Optic disc area is a circular and quite distinct bright region in a digital colour fundus image. This paper provides new method to get an optic disc area automatically. Our method works with some steps. In the first step, the red channel is extracted. Then, the optic disc area is localised using circular average filter to detect the candidate of optic disc point. This point is used to establish the region of interest (ROI). After creating the ROI, multiple bottom hat transformation is employed to detect the blood vessels. To remove blood vessel regions, the result of multiple bottom hat transformation is added to ROI image. Before segmentation is performed, image smoothing is done using average filter. The method is tested on forty different optical disc images from DRION database. Our proposed method achieves an average level of sensitivity, specificity and accuracy of 96.12%, 94.36 % and 94.71%, respectively. This indicates that the proposed algorithm successfully detects and segments the optic disc area and is able to be implemented as measurement for the projection line to get the centre of the fovea.