Shuzlina Abdul Rahman

Towards emotional control recognition through handwriting using fuzzy inference

Emotion control is one of personality characteristics that can be detected through handwriting or graphology. One of the advantages is it may help the counselor that has difficulties in identifying the emotion of their counselee. This study is to explore the fuzzy technique for feature extraction in handwriting and then identify the emotion of person. This study uses baseline or slope of the handwriting in determining the level of emotion control whether it is very low, low, medium, high or very high, through Mamdani inference.

Online Slant Identification Algorithm Using Vector Rules

Signatures are among the most widely accepted personal attributes for identity verification. There are a lot of features that can be discovered in signature which are either dynamic or static features type. An algorithm needs to be designed to extract these signature features. Online system uses pressure sensitive tablets to capture signature of individual as they sign thus analysis can be done directly and immediately. This research explored slant feature algorithm since signature is usually slanted due to the mechanism of handwriting and the human personality. The proposed algorithm are used to formulate the Signature Extraction Features System (SEFS) which provides a set of tools that allow the users to extract slant features in signature automatically for analysis purposes. Twenty individuals from different background are randomly selected to have their signature taken. Their signatures are captured on a tablet and the SEFS would than gather and store the raw data. The image of the signature that is created by the SEFS would be used as samples for the questionnaire to identify the features of slant, where the questionnaires are given to human expert for evaluation. The results from the SEFS are compared with the result from the questionnaire. Results produced by the algorithm for slant extraction shows 85% identical answers compared to the outcome by human expert. These show that the algorithm proposed are promising for further exploration.

Data mining framework for protein function prediction

Determining the functions of uncharacterized proteins from sequences remains a challenge despite the growth of the number of prediction methods. This is due to the nature of the inherent limitations of current tools and databases and the ambiguity of the function definition. Additionally, standard methods of functional assignment involve sequence alignment to a gene function often fail to find the significant matches. This paper proposes a framework of machine learning method in predicting protein function irrespective of sequence similarity. The framework aims to provide a workflow on predicting protein function that combines both data mining and machine learning algorithms. Three main components are involved: pre-processing, model development and testing & evaluation. The study is expected to create a new method on feature selection processes towards predicting protein functional classes in addition to complementing the existing conventional method of functional assignment.

Plant Selection System

Expert in agriculture field is in demand, so storing their knowledge would be useful. This research is to analyze the information in agriculture by constructing rules in evaluating the potential plants based on soil suitability and also marketability, yields, consumption or budget. The expertpsilas knowledge was extracted from an agriculture officer and represented in the IF-THEN rules using confident factor. The suggestion is created based on the percentage that was above 80%. In conclusion, the system can give explanations about potential plants, the financial and marketing support and fertilization. It is recommended to apply hybrid expert system and others in future, in order to get even better and accurate results.

Towards applying associative classifier for genetic variants

With the availability of biological data and the power of sharing, it produces many opportunities for computer scientists to perform researches in bioinformatics. Generally the researches propose methods for different tasks, mainly to develop algorithms in diagnosing and identification of diseases. One of the primary studies that relevant to health and diseases is genome wide association studies (GWAS). Normally the studies are conducted in different populations to replicate the risk loci of specific disease and the number of groups are keep on progressing, including those from Asian country. Computer scientists should be involved in GWAS due to certain problems and the complexity of the processes involved. The problems and past studies related to GWAS are presented in this paper.

Fuzzy water dispersal controller using Sugeno approach

Controlling the amount of water in maintaining lawn health and beauty is a main topic in horticulture. A reliable controller is needed to control the amount of water to disperse as to ensure the soil has enough moisture adequacy level. This research explores the use of fuzzy logic for the controlling of water dispersal. The performances of fuzzy water dispersal controller (FuziWDC) was measured based on a significant set of common Bermuda turfgrass. An improved Sugeno inferencing for the task of water dispersal controller is presented that considered both evapotranspiration (ET), tensiometer variable as opposed to the earlier work. A comparison of the output is being performed to evaluate Sugeno with conventional system. The result shows FuziWDC has performed better than the conventional technique based on the lower annual average water usage for the whole year recorded.

Weak Classifiers Performance Measure in Handling Noisy Clinical Trial Data

Most research concluded that machine learning performance is better when dealing with cleaned dataset compared to dirty dataset. In this paper, we experimented three weak or base machine learning classifiers: Decision Table, Naive Bayes and k-Nearest Neighbor to see their performance on real-world, noisy and messy clinical trial dataset rather than employing beautifully designed dataset. We involved the clinical trial data scientist in leading us to a better data analysis exploration and enhancing the performance result evaluation. The classifiers performances were analyzed using Accuracy and Receiver Operating Characteristic (ROC), supported with sensitivity, specificity and precision values which resulted to contradiction of conclusion made by previous research. We employed pre-processing techniques such as interquartile range technique to remove the outliers and mean imputation to handle missing values and these techniques resulted to; all three classifiers work better in dirty dataset compared to imputed and clean dataset by showing highest accuracy and ROC measure. Decision Table turns out to be the best classifier when dealing with real-world noisy clinical trial.

Review of Frequent Itemsets Mining in High Dimensional Dataset

Nowadays, there are abundant of big data collection and to understand its patterns would need a thorough analysis. Analyzing big data would depend highly on the purpose and the tasks involved would be various. One of the significant tasks is frequent itemsets mining and the strategy has been evolved in many ways in order to improve the efficiency and effectiveness of the mining process. In this paper, we briefly reviewed mining frequent itemsets algorithms from year 1998 until year 2013 that focus on maximal and closed frequent itemsets. We discussed these algorithms based on three main areas namely: the searching strategy, space reduction method, and data representation. These three main areas are concluded as the optimization strategy and are designed to improve the efficiency and scalability using a different approach in different areas to adapt to numerous growth of the dataset. This work is beneficial for researchers in designing and enhancing the algorithm based on their own purposes.