Noor Azilah Muda

SOCIFS feature selection framework for handwritten authorship

The uniqueness of shape and style of handwriting can be used to identify the significant features in confirming the author of writing. This paper is meant to propose a novel feature selection framework for Swarm Optimized and Computationally Inexpensive Floating Selection SOCIFS, by exploring existing feature selection frameworks, and compare the performance of proposed feature selection framework against various feature selection methods in Writer Identification in order to find the most significant features. The promising applicability of the proposed framework has been demonstrated in the result and worth to receive further exploration in identifying the handwritten authorship.

PSO and Computationally Inexpensive Sequential Forward Floating Selection in acquiring significant features for handwritten authorship

The uniqueness of shape and style of handwriting can be used to identify the significant features in confirming the author of writing. Acquiring these significant features leads to an important research in Writer Identification domain. This paper is meant to explore the usage of feature selection in Writer Identification in order to find the most significant features. This paper proposes a hybrid feature selection method of Particle Swarm Optimization and Computationally Inexpensive Sequential Forward Floating Selection for Writer Identification. The promising applicability of the proposed method has been demonstrated and worth to receive further exploration in identifying the handwritten authorship.

Immune ant swarm optimization for optimum rough reducts generation

Ant Swarm Optimization refers to the hybridization of Particle Swarm Optimization PSO and Ant Colony Optimization ACO algorithms to enhance optimization performance. It is used in rough reducts calculation for identifying optimally significant attributes set. This paper proposes a hybrid ant swarm optimization algorithm by using immunity to discover better fitness value in optimizing rough reducts set. By integrating PSO with ACO, it will enhance the ability of PSO when updating its local search upon quality solution as the number of generations is increased. Unlike the conventional PSO/ACO algorithm, proposed Immune ant swarm algorithm aims to preserve global search convergence of PSO when reaching the optimum especially under the high dimension situation of optimization with small population size. By combining PSO with ACO algorithms and embedding immune approach, the approach is expected to be able to generate better optimal rough reducts, where PSO algorithm performs the global exploration which can effectively reach the optimal or near optimal solution to increase fitness value as compared to the past research in optimization of attribute reduction. This research is also to enhance the optimization ability by defining a suitable fitness function with immunity process to increase the competency in attribute reduction and has shown improvement of the classification accuracy with its generated reducts in solving NP-Hard problem. The proposed algorithm has shown promising experimental results in obtaining optimal reducts when tested on 12 common benchmark datasets. Result for rough reducts and fitness value performance has been discussed and briefly explored in order to identify the best solution. The experimental analysis on the initial results of IASORR has been proven to offer a better quality algorithm and to maintain PSOs performance, which are also encouraging in t-test analysis, for most of the tested datasets.

Historic Document Image De-noising Using Principal Component Analysis (PCA) and Local Pixel Grouping (LPG)

In this paper, an approach of principal component analysis (PCA) with local pixel grouping (LPG) is used to de-noising the noisy historical document image. This technique ensures the preservation of historic document image local structure. This is due to block matching based LPG which carries out classification to allow only the sample blocks with similar contents used in the calculation for PCA transform estimation. Such an LPG procedure ensures that the image local features can be well preserved after the noise removing process in the PCA domain. The LPG-PCA de-noising procedure will repeat one more times with adaptively adjusted noise level to further improve the performance of de-noising the historic document image. The experiment results show that LPG-PCA model has good results in de-noising historical document image.

A Comparative Study of Tree-based Structure Methods for Handwriting Identification

Handwriting Identification is a process to determine the author of the writing and it involves some of process. Classification process is a final stage of Handwriting Identification process where it will analyze the classification accuracy and based on the number of features selected. In this study, classification process was conducted using various tree-based structure methods. Tree-based structure method is one of the classification methods where it is able to generate a compact subset of non-redundant features and hence improves interpretability and generalization. However its focus is still limited especially in Writer Identification domain. Several of tree-based structure selected and performed using image dataset from IAM Handwriting Database. The results also analyze and compared of each methods of Writer Identification. Random Forest Tree classifier gives the best result with the highest percentage of accuracy followed by J48, Random Tree, REP Tree and Decision Stump.

Improving ant swarm optimization with embedded vaccination for optimum reducts generation

Ant Swarm Optimization refers to the hybridization of Particle Swarm Optimization (PSO) and Ant Colony Optimization (ACO) algorithms to enhance optimization performance. It is used in rough reducts calculation for identifying optimally significant attributes set. This paper proposes a hybrid ant swarm optimization algorithm by using immunity to discover better fitness value in optimizing rough reducts set. Unlike a conventional PSO/ACO algorithm, this hybrid algorithm shows improvement of the classification accuracy in its generated rough reducts to solve NP-Hard problem. This paper has evaluated the immune algorithm in 12 common benchmark dataset to evaluate the performance of rough reducts-based on attribute reduction. The results show that immune ant swarm algorithm is very competitive in terms of fitness value, number of iterations, and classification accuracy to produce a better optimization technique and more accurate results in rough reducts generation. The results also show that immune ant swarm optimization provides a slight increase in accuracy when compared to the differential evolution variant.

Recognizing Music Features Pattern Using Modified Negative Selection Algorithm for Songs Genre Classification

Previous studies have proven that imitating the mechanism of recognizing alien cells is beneficial and provides so many solutions to the pattern recognition related problems. These efforts emulate the human immune system in recognizing the cells by considering every essential component or features of the subjects. In this research, the focus is on analyzing the music features patterns to recognize various songs genres by emphasizing the features from the artists’ voices, the melody of the music and even the sounds of the musical instruments used. Three fundamental music contents are investigated which are timbre, rhythm, and pitch based features. The main objective of this research is to recognize the music features from different genres using the modified negative selection algorithm fundamental procedures that are the censoring and monitoring modules. The results of the experimental works are remarkable and are comparable to previous works in the music recognition and classification works. In this highlight, stages of music recognition are emphasized where feature extraction, feature selection, and feature classification processes are explained. Comparison of performances between proposed algorithm and other classification technique are discussed.

Handling a Small Dataset Problem in Prediction Model by employ Artificial Data Generation Approach: A Review

The emerging era of big data for past few years has led to large and complex data which needed faster and better decision making. However, the small dataset problems still arise in a certain area which causes analysis and decision are hard to make. In order to build a prediction model, a large sample is required as a training sample of the model. Small dataset is insufficient to produce an accurate prediction model. This paper will review an artificial data generation approach as one of the solution to solve the small dataset problem.

Tree-base Structure for Feature Selection in Writer Identification

Handwriting is individualistic where it presents various types of features represent the writer’s characteristics. Not all the features are relevant for Writer Identification (WI) process and some are irrelevant. Removing these irrelevant features called as feature selection process. Feature selection select only the importance features and can improve the classification accuracy. This chapter investigated feature selection process using tree-base structure method in WI domain. Tree-base structure method able to generate a compact subset of non-redundant features and hence improves interpretability and generalization. Random forest (RF) of tree-base structure method is used for feature selection method in WI. An experiment is carried out using image dataset from IAM Hand-writing Database. The results show that RF tree successively selects the most significant features and gives good classification performance as well.

Rotation analysis of moment invariant for 2D and 3D shape representation for molecular structure of ATS drugs

Drug abuse among the people around the world is getting serious and increase nowadays. An Amphetamine-Type Stimulant (ATS) drug is one of the popular drugs in the world. This kind of drug is a combination of two types of drugs; amphetamine and ecstasy. In this project, we are focusing on two different techniques used to compare the result between 2D and 3D ATS drug molecule. Besides that, this project is also looking for the best technique to produce 2D and 3D images. United Moment Invariant (UMI) and Hough Transform (HT) are the feature extraction methods that will be used to extract data from both 2D and 3D molecular structure of the ATS drugs. Later, the extracted data will be processed into WEKA for classification accuracies for both molecular structures. As a conclusion, the MI and the HT technique will definitely provide different results, but both techniques will perform well with the ATS molecular structure.