M. N. Vijayalakshmi

Mining of student academic evaluation records in higher education

The various data mining techniques like classification, clustering and relationship mining can be applied on educational data to predict the performance of a student in the examination and bring out betterment in his academic performance. Rule based classification techniques can be used to predict the result of the students in the final semester based on the marks obtained by them in the previous semesters. Decision table and One R rule algorithms are used here to predict the result of the fifth semester student based on the marks obtained by the students in the previous four semesters. The accuracy of the algorithms is analyzed by comparing the prediction given by miner, the algorithm and the result obtained by the students in the fifth semester examination. The performances of the algorithms are analyzed using the true positive, false positive values and the ROC curves.

Analysis of wild fire behaviour in wild conservation area using image data mining

Forest fire is a major environmental issue creating ecological damage. Fire detection is a key element for controlling such incident. As per the forest survey of India 19.27% or 63.3 million hectare of the Indian land has been classified as forest area, of which 38 million ha alone are hoarded with resources in great quantity (top density above 40%).There are many fire detection algorithms available, each one of it has its own approach of predicting fire. The proposed work processes the satellite images based on its intensity levels to find out the fire affected region (hot spots). In order to detect hot spots agglomerative hierarchical clustering algorithm is used and the direction of the fire spread regions are plotted based on the clusters obtained by the algorithm for the given input image. The implementation is based on RGB values of pixels of an image. The algorithms efficiency is relatively high when it is applied on forest fire images.

Inference of Naïve Baye's Technique on Student Assessment Data

The main objective of any educational institute is to provide quality education to students and produce qualified students to the community. This can be achieved only when the institutions are capable of predicting the student’s behavior, their attitude towards studies and also the outcome of their result in the forth coming examinations. This can be achieved through various data mining techniques like classification, clustering and rule based mining. Classification techniques like decision trees, Bayesian network and neural networks can be used to predict the student’s outcome in the examination based on their attendance percentage, their marks in the internal examination, the historical data available in the form of their previously scored percentage etc. Bayesian classification technique is used to predict the student’s outcome in the university examination based on the marks obtained by them in the internal examination. Bayes classification is used to predict the result of the student on an individual basis which has helped the tutor to identify the weak students in each subject. This result has helped the tutors to concentrate on those weak students and bring out better results. This prediction will also help the institution to reduce the drop out ratio and produce better results.

Analysis of Square and Circular Diaphragms for a MEMS Pressure Sensor Using a Data Mining Tool

Micro electromechanical systems (MEMS) are small integrated 3 dimensional devices having structures ranging from sub micron level to millimeter level that combine electrical and mechanical components on a single substrate. They find various applications in mechanical, optical and biological fields due to a reduction in size, the excellent mechanical properties of silicon, extension of the well developed designing, fabrication and packaging processes of the Integrated Circuit (IC) industry, cost effectiveness etc. The special feature of a smart MEMS sensor is the ability to integrate sensing, controlling and actuating functions on a single chip. In this paper various factors which play a critical role in the performance of a MEMS pressure sensor diaphragm are discussed. Various parameters like side length, radius, and maximum stress induced in a square and circular diaphragm are calculated for a pressure range of 0.1 to 1MPa. The effects of variations in the applied pressure on the geometry of the sensing element and the relationship between applied pressure, the induced stress and the dimensions of the micro sensing diaphragm are discussed using the WEKA data mining tool.

Optimization of circuit partitioning in VLSI through classification algorithms

In order to build complex digital logic circuits it is often essential to sub-divide multi million transistors design into manageable pieces. Circuit partitioning in VLSI, is one of the major area of research. There are many existing diverse algorithm to partition the circuit into sub circuits. This paper aims at circuit partitioning using two classification algorithms Decision Tree Algorithm and K-Nearest Neighbors Algorithm. These two algorithms were tested on a 3-bit Priority Encoder and a 4×2 SRAM sample circuits and implemented using VHDL. The tested result shows that the K-Nearest Neighbor algorithm yields better subcircuits than the Decision Tree Algorithm.

Efficiency of Multi-instance Learning in Educational Data Mining

Educational data mining (EDM) is one of the emerging technologies in recent years. The various changes in the process of teaching and learning have brought in a lot of challenges to the stakeholders to understand the learners toward the different methods of teaching and the way they perform in various teaching environments. This chapter is an application of Baker’s taxonomy in an educational dataset to predict course outcome of the learners during the middle of the course. The experiment is conducted using different single and multi-instance-based learning algorithms. The efficiency of the single and multi-instance learning algorithms was measured using the accuracy rates and the time taken to build the model. In single instance algorithm, decision stump tree was found very effective and in multi-instance learning, the Simple MI method was found very effective. The precision of the instance-based learning algorithms is calculated using Wilcoxon rank method, and multi-instance learning algorithm is found to be more accurate than the single instance learning techniques.

An efficient and optimal clustering algorithm for real-time forest fire prediction with

Grouping of forest fire images into meaningful categories to reveal useful information is a challenging task. In order to overcome this challenge, data mining techniques can be used with wireless sensor network which can detect and forecast forest fire more promptly than the satellite-based detection approach. This paper proposes an efficient image clustering algorithm using real time data for predicting of the occurrence of forest fire, with a new mechanism for secure information transmission in wireless sensor networks by minimizing the threat attacks caused by malicious nodes in wireless sensor networks.