Search

Histological staining facilitates the microscopic study of diseased tissues in clinical investigations. Staining by fluorescent dyes enhances the tissue contrast, highlighting the important features. Haematoxylin and Eosin staining protocol is the most common in the histopathological analysis of cells and tissues in cancer diagnosis. HE staining provides a visual representation of tissue abnormalities, distinguishing cell nuclei and acidophilic structures. Stain unmixing is an equally vital procedure which decomposes the multi-stained image into individual stain components. These components are essential for visual examination of the interaction between the dye and the specific tissues. This manuscript put forwards an original cell destaining approach based on a three factor Nonnegative Matrix Factorization which separates the H and E components from the HE stained image. The experimental results with standard datasets and performance metrics demonstrate the robustness of the method yielding better PSNR and SSIM values for stain separation compared to similar works in this context. The PSNR values around 30 dB are achieved by the proposed unmixing approach which is more than 10 dB compared to the benchmark approaches. Similarly, the SSIM values are obtained in the range 0.42–0.73, signifying strong structural elements. The proposed method can be further extended to study the interaction of other activators in the wound repair process under cancer progression.

Video event recognition plays an important role in the various research fields particularly in surveillance detection system. In the existing system it is done by deep hierarchical context model which utilizes several contextual data which contains various context information at level of feature, semantic and priority for the video recognition process. However, this research method might perform low with increased volume of videos and might be failed to predict the events accurately with less interrelation contextual features. The standstill challenges are solved via improved hybridized deep structured model. The three primary features of contextual data are to discriminate resultant neighborhood. Here hybrid textual perceptual descriptor and concept-based attribute extraction is performed for accurate recognition of video events. These extracted interaction context features are grouped by using improved K means algorithm. In addition, improved deep structured model that combines convolutional neural networks and conditional random fields are developed for learning middle level representations and mingle the bottom feature level, mid-semantic and top-level meanings for the identification of incidents. This proposed research method is evaluated by using VIRAT data set whose simulation analysis is performed using Matlab simulation toolkit. The overall evaluation of the proposed research method proves that the suggested method can provide better output in terms of accurate recognition of events.

In this propose technique for colorizing photographs, the shading variant of a given dim scale picture is taken from it. The technique does not depend upon human data, and is totally modified. It doesn't depend upon division, writing or complex picture taking care of strategies. It relies upon setting up a direct classifier applying back inducing over a preparation set of shading and looking at diminish scale pictures. The classifier predicts the shade of a pixel reliant on the diminish level of the pixels enveloping it. This little fix gets a close by surface. To save the space for the pointer little, the shades are reduced using Self Organizing Maps. This lessening conveys a little plan of Chroma regards with enough assortment as to make extraordinary approximations for all tones in the planning set.

Environment influences plants in various forms of stress that affects plant’s growth and development. Biotic Stress causes damage to plants with the help of living organisms, pests, and pathogens. This leads to a huge loss of agricultural productivity and sustainability that has a serious impact on the world economy. Thus, at the outset of growth, it is highly desirable to automate the process of disease identification and diagnosis. At present, computer vision using deep learning methods have achieved considerable success with object detection tasks. The goal of this work is to develop an innovative, ground-breaking framework for the detection of diseases based on a deep convolution neural network through the transfer learning technique. For this work, initially, we collected a dataset that contains images of 1,014 Betelvine leaves with healthy and diseased categories in the cultivated betel vine crops and annotated manually for training. Mask-RCNN instance segmentation model is adopted for detecting Betelvine diseased leaves effectively. Eventually, the trained model achieved mean Average Precision (mAP) of 84%, using an updated Mask-RCNN model by fine-tuning the parameters with a swish as activation function and RAdam as the optimizer.

The advancement of the internet coined a new era for inventions. Internet of Things (IoT) is one such example. IoT is being applied in all sectors such as healthcare, automobile, retail industry etc. Out of these, Human Activity Recognition (HAR) has taken much attention in IoT applications. The prediction of human activity efficiently adds multiple advantages in many fields. This research paper proposes a HAR system using the wearable sensor. The performance of this system is analyzed using four publicly available datasets that are collected in a real-time environment. Five machine learning algorithms namely Decision tree (DT), Random Forest (RF), Logistics Regression (LR), K-Nearest Neighbor (kNN), and Support Vector Machine (SVM) are compared in terms of recognition of human activities. Out of this SVM responded well on all four datasets with the accuracy of 77%, 99%, 98%, and 99% respectively. With the support of four datasets, the obtained results proved that the performance of the proposed method is better for human activity recognition.

Activity recognition(AR) is a popular subject of research in the recent past. Recognition of activities performed by human beings, enables the addressing of challenges posed by many real-world applications such as health monitoring, providing security etc. Segmentation plays a vital role in AR. This paper evaluates the efficiency of Area-Based Segmentation using different performance measures. Area-Based segmentation was proposed in our earlier research work. The evaluation of the Area-Based segmentation technique is conducted on four real world datasets viz. Aruba17, Shib010, HH102, and HH113 comprising of data pertaining to an individual, living in the test bed home. Machine learning classifiers, SVM-R, SVM-P, NB and KNN are adopted to validate the performance of Area-Based segmentation. Amongst the four chosen classification algorithms SVM-R exhibits better in all the four datasets. Area-Based segmentation recognise the four test bed activities with accuracies of 0.74, 0.98, 0.66, and 0.99 respectively. The results reveal that Area based segmentation can efficiently segment sensor data stream which aids in accurate recognition of smart home activities.

The electrical activity of the human heart is measured by the vital bio medical signal called ECG. This electrocardiogram is employed as a crucial source to gather the diagnostic information of a patient’s cardiopathy. The monitoring function of cardiac disease is diagnosed by documenting and handling the electrocardiogram (ECG) impulses. In the recent years many research has been done and developing an enhanced method to identify the risk in the patient’s body condition by processing and analysing the ECG signal. This analysis of the signal helps to find the cardiac abnormalities, arrhythmias, and many other heart problems. ECG signal is processed to detect the variability in heart rhythm; heart rate variability is calculated based on the time interval between heart beats. Heart Rate Variability HRV is measured by the variation in the beat to beat interval. The Heart rate Variability (HRV) is an essential aspect to diagnose the properties of the heart. Recent development enhances the potential with the aid of non-linear metrics in reference point with feature selection. In this paper, the fundamental elements are taken from the ECG signal for feature selection process where Bat algorithm is employed for feature selection to predict the best feature and presented to the classifier for accurate classification. The popular machine learning algorithm ELM is taken for classification, integrated with evolutionary algorithm named Self- Adaptive Differential Evolution Extreme Learning Machine SADEELM to improve the reliability of classification. It combines Effective Fuzzy Kohonen clustering network (EFKCN) to be able to increase the accuracy of the effect for HRV transmission classification. Hence, it is observed that the experiment carried out unveils that the precision is improved by the SADE-ELM method and concurrently optimizes the computation time.

This work uses crowd sourcing to examine the benchmark datasets of the specified areas using data-mining and machine learning algorithms. The proposed system will allow the users to update their local places by providing the location and gives the best shortest path, time required to travel, minimum cost and suggests mode of transport in shared basis. It suggests better routes available and the resources (public transports/two-wheelers/cars, etc.) through which we achieve cost efficiency. In case of an emergency situation, we also suggest the alternate routes and modes of transport, so that the implementation will also be time efficient. This implementation can be extended by user ideas with the data analysis result to the government by suggesting alternative modes such as (bridges, subways, temporary passages) to the accident-prone zones and school zones where the traffic intensity is usually high.

Unmanned Aircraft Systems (UAS) have now been used in various fields including emergency services to army drones in the present day. Hence there exists a requisite for developing an automation to avoid any obstacle that it encounters. Since these aircrafts are almost present in most places, there is also a need for noise reduction for public health consideration and as well as keeping the target clueless while vigilance in enemy’s scope. In this paper, we implement different types of machine learning algorithms mainly different regressors to predict the airfoil sound pressure which can be further integrated and used for avoiding obstacles on the way of UAS with denoising. The proposed work dishes out one of the simple and fairly accurate models for the prediction of the sound pressure generated by UAS, using which any mechanical methods can be applied for its reduction. The algorithms are experimented with a sample data set and the comparison between these algorithms has been made on the same data. Random Forest algorithm performs better than the other specified algorithms with an accuracy of around 83%.

Analysis of micro array data for disease prediction is an arduous machine learning job. Due to its high proportions and low samples, the computational risk is high. To increase the efficiency of the classifier a Signal-to-Noise Ratio algorithm can be used to get the most relevant genes. Then to these selected gene datasets, an ensemble-based classification algorithm is applied to get an efficient classifier for future prediction. The ensemble-based algorithm is an optimal combination of a set of learner algorithms. Our research focuses on 3 main classification algorithms namely SVM, KNN and Decision Tree which increases the performance and give confidence to the result. The ensemble is done using majority voting method which outperforms the base classifier.

Ambient monitoring is a much discussed area in the domain of smart home research. Ambient monitoring system supports and encourages the elders to live independently. In this paper, we deliberate upon the framework of an ambient monitoring system for elders. The necessity of the smart home system for elders, the role of activity recognition in a smart home system and influence of the segmentation method in activity recognition are discussed. In this work, a new segmentation method called area-based segmentation using optimal change point detection is proposed. This segmentation method is implemented and results are analysed by using real sensor data which is collected from smart home test bed. Set of features are extracted from the segmented data, and the activities are classified using Naive Bayes, kNN and SVM classifiers. This research work gives an insight to the researchers into the application of activity recognition in smart homes.

Social Media now plays a significant role in shaping and impacting our attitude, behaviour and cultural aspects of our life. The general mood or opinion about a product or a subject matter can be derived by monitoring social media activities of individuals pertaining to the product or subject matter. Tracking these and analysing them helps in a more targeted advertising campaigns for example. The proposed work explores Sentiment Analysis study performed on a sample of Twitter posts on the subject of Make in India, a Govt. of India (GOI) initiative to hasten and support growth in the Indian manufacturing sector. This paper identifies the overall sentiment towards Make in India initiative and also identifies the sentiment towards various sectors which are part of the initiative. Additionally, the sentiment and sector classification are done separately based on the type of Twitter users, namely a personal user account or business/organizations. The latter type of user has a tendency to deeply influence and skew the sentiment results and therefore it is important to separate out this point of view during the analysis of public’s sentiment on the subject. The analysis is performed using machine learning approach. Unigram and Bigram feature extraction techniques are used and their correlation is calculated using Count Vectorizer and Term Frequency – Inverse Document Frequency (TF-IDF). The prediction is done using supervised machine learning algorithms and uses R Studio for the analysis.

Social Network (SN) is an online platform broadly used as communication tool by millions of users in order to build social relationships with others for knowledge point of view, career purposes and many more. Social Networks such as Twitter, Facebook, and LinkedIn have become the most leading tools on the web. Spam, floods the Internet with many copies of the same message and it can be manifest in numerous ways, it includes bulk messages, malicious links, fake friends, fraudulent reviews and personally identifiable information. The aim of this paper is to classify the tweets into spam and non-spam using Machine Learning and which will give the best results.

Sentiment analysis with opinion mining is an aspect of Text Analytics under the branch of Data Mining. We might be surprised to know that it plays an influential role in our daily lives. For instance, the decisions made over picking up a tooth paste, or in buying a car. Many aspects of human behaviour are exploited by targeted advertising, the directions to which are arrived at based on sentiment analysis techniques. The decisions are made over a large array of data called dataset. The proposed work involves analysis of tweets on the hashtag Make in India, an initiative by Mr. Narendra Modi, Prime Minister of India. Here the tweets act as the dataset. The tweets are collected, cleaned, filtered which is known as pre-processing. Then the features are extracted and selected for the analysis. The classification, evaluation and visualization are carried out using machine learning techniques. ‘R Studio’ is used for implementation and analysis in this work.

Data mining is the process that is used to extract the meaningful information from the large size dataset. The effective dataset utilization depends on the proper classification of outliers. The main objective of the outlier detection is to extract the abnormal data with inconsistency. The information collection from the different mechanisms is uncertain in nature. The data uncertainty causes the knowledge imperfections namely, vagueness and indiscernibility. The proposed research work implements an efficient fuzzy–rough set classifier for an outlier detection with less computational complexity. The fuzzy logic utilization and the fix of abnormal data easily determine the outliers in the large size database. The proposed classifier is compared with the existing classification methods with performance parameters like average running time, average execution time, execution time, false negative, false positive, true negative, true positive, precision, recall, and accuracy. The comparative analysis depicts the effectiveness of classifier in outlier detection.

Classification of different tumor type are of great significance in problems cancer prediction. Choosing the most relevant qualities from huge microarray expression is very important. It is a most explored subject in bioinformatics because of its hugeness to move forward humans understanding of inherent causing cancer mechanism. In this paper, we aim to classify leukaemia cells. Our approach relies on selecting the predominant features from the dataset and classifying it using classification algorithm, Support Vector Machine (SVM). As the dataset is very large we need to reduce the size before classification, to decrease the computation time and increase the accuracy of the classifier. SVM-RFE removes only most irrelevant gene in each iteration. This algorithm does not differentiate the correlated genes. So before applying SVM-RFE for gene selection, a correlation-based feature selection is applied using FCBF (Fast Correlation-Based Filter) to select the most prominent not correlated genes. The resultant classifier generated using this set of genes yields more accurate result and reduce the computational time.

This article presents an overview of hyperspectral image analysis and processing techniques based on remote sensing. Image analysis methods will be explained in detail. A general framework is presented for working with hyperspectral imagery, including removal of atmospheric effects. Due to large dimensionality of the feature space, hyperspectral data poses a challenge to image interpretation in the following ways: 1) need of calibration of data2) redundancy in information and 3) high volume data. Hence, a brief discussion on dimensionality reduction will also be presented in this review.

Thresholding methods are important in any image processing technique. It is used to convert the input gray level image to binary image. Hence, they are the essential preprocessing methods. The present paper proposes various methods for thresholding based on second order statistics (CTH, ETH, HTH, ATH). The proposed four methods are applied on various finger print images and compared with various first order statistics based methods (max, min, (max-min)/2, standard deviation, mean). The results clearly indicate the efficiency of the proposed algorithms.

Digital watermarking is a technique developed to help copyright protection of Image, audio & video files. It has been spurred by the growth of the Internet, and the ease of copying. This has led to an unprecedented numbers in copyright infringements. The digital audio & video watermarking is concerned with embedding data in a signal the watermark data can be accurately detected. Watermarking algorithm based on Empirical Mode Decomposition (EMD) is introduced. The signal is divided into frames and each one is decomposed adaptively by Empirical Mode Decomposition into intrinsic oscillatory components. The data embedding rate of EMD algorithm is 46.9-50.3 b/s; the robustness of hidden watermark is identified using Mp3 compression, additive noise, and quantization, cropping, filtering and resampling.

A data warehouse is a collection of data gathered and organized so that it can easily be analyzed, extracted, synthesized and also be used for the purpose of further understanding data. Peer to Peer networks are used for distribution and sharing of documents. In traditional techniques, when aggregate functions like average, sum and count are encountered, the aggregate operation is performed by considering all the nodes and tuples, which reduces the efficiency of the query processing system. Exact solutions can be time consuming and difficult to implement, given the distributed and dynamic nature of P2P networks. The problem is overcome in this project by selecting random peers and random tuples from P2P networks and then performing the aggregation operation, thus the speed is increased and latency is reduced. Though accuracy is compromised to small extent, efficiency is achieved. Thus, this kind of approximate query processing will be beneficial to the areas where efficiency plays a main role than accuracy. Adaptive Hybrid approach based on random walk is used to achieve the efficiency in the performance of aggregation operation.

This paper proposes a novel method, a Fuzzy Feature Match (FFM) based on a controlled triangle feature set inscribed in a rhombus to match the deformed fingerprints. The fingerprint is represented by the fuzzy feature set. The fuzzy features set similarity is used to analyze the similarity among fingerprints. A fuzzy similarity for two triangles is introduced and then a similarity vector including the triangle-level similarities for all triangles and the rhombus in two fingerprints are constructed. Accordingly, a similarity vector pair is defined to illustrate the similarities between two fingerprints. The FFM method shows the similarity vector pair to a normalized value which quantifies the overall image to image similarity. The algorithm has been evaluated with Noorul Islam (NI) university staff members' fingerprint database. Experimental results confirm that the proposed FFM based on the local triangle feature set inscribed in a rhombus is a reliable and effective algorithm for fingerprint matching with nonlinear distortions.

We propose a novel vision-based lane departure warning algorithm. The initial step of lane departure warning system is lane detection. In this paper, we present a lane detection method based on template matching. Kalman filtering is used to track the detected lanes. To make the algorithm robust to shadows, inpainting technique is done on regions with shadows. The curved lanes are handled with the help of Bezier spline fitting. Our algorithm determines the lane departure based on vanishing point margins. Experimental results show that the proposed method performs better than Caltech lane detection algorithm.

In data mining, clustering is one of the most significant task, and has been widely used in pattern recognition and image processing. One of the tradition and most widely used clustering algorithm is k-Means clustering algorithm, but this algorithm fails to find structural similarity in the data or if the data is non-linear. Spectral clustering is a graph clustering method in which the nodes are clustered and useful if the data is non-linear and it finds clusters of different shapes. A spectral graph is constructed based on the affinity matrix or similarity matrix and the graph cut is found using Laplacian matrix. Traditional spectral clustering use Gaussian kernel function to construct a spectral graph. In this paper we implement PCA based similarity measure for graph construction and generated different Laplacian graphs for spectral clustering. In PCA based similarity measure, the similarity measure based on eigenvalues and its eigenvectors is used for building the graph and we study the efficiency of two types of Laplacian graph matrices. This graph is then clustered using spectral clustering algorithm. Effect of PCA similarity measure is analyzed on two types of Laplacian graphs i.e., un-normalized Laplacian and normalized Laplacian. The outcome shows accurate result of PCA measure on these two Laplacian graphs. It predicts perfect clustering of non-linear data. This spectral clustering is widely used in image processing.

Support Vector Machine (SVM) [2], is most widely popular learning algorithm used for classification of large dataset. Our project aims to generate a classifier for breast cancer genes microarray by using modified-SVM-RFE algorithm. This breast cancer microarray contains a large number of genes and its expression, so it necessary to reduce the number of genes before applying for classification. So the most efficient algorithm that can be applied for classification of microarray is SVM-RFE [3][8], which is an embedded method, which performs backward single gene elimination as well as classification of the dataset. A new modified algorithm is proposed with less computation over SVM-RFE. SVM-RFE generates the rank of the features and eliminates one lowest rank irrelevant feature, in each iteration. Since our microarray contains 47,294 genes its very computational overhead to reduce the dimension. So the modified algorithm which removes more than one irrelevant genes in single iteration of SVM-RFE algorithm. And also this algorithm only removes irrelevant gene, it does not remove the correlated genes. So before applying SVM-RFE, our research focuses on finding out the correlated genes and extracting a new gene from the two, and then apply SVM-RFE on the new set of genes. So our proposed method is Correlation based Support Vector Machine Recursive Multiple Feature Elimination (CSVM-RMFE) algorithm which first extracts a new genes from two correlated genes called virtual gene and then apply SVM-RMFE to generate a classifier. This SVM-RMFE algorithm eliminate multiple feature so that the classification time can be reduced and its accuracy can be increased.

In the classification of the heart disease data set a high dimensional data set is used in the pre processing stage of data mining process. This raw dataset consist of redundant and inconsistent data thereby increasing the search space and storage of the data. To achieve the classification accuracy we need to remove the redundant and the irrelevant data present. The dimensionality reduction technique is used to compress the high dimensional data to lower dimensional data with some constraints. A framework is integrated for the easy prediction of the heart disease. The framework is created by using the principal component analysis (PCA) to extract the features and mathematical model is computed to select the relevant features using the relevant constraint. This proposed work helps in improving the efficiency, accuracy and speed of the process. This can be applied in the applications such as information retrieval, image processing and pattern matching.

Pedestrian Detection (PD) is an active research area for improving road safety. Most of the existing PD system does not meet the demanded performance. This paper presents a working PD system which improves performance. The system uses CENTRIST feature extractor and the linear Support Vector Machine (SVM) for training and detection of pedestrian. CENTRIST is very easy to compute without any preprocessing and normalization that makes it suitable for on-board system. During the training procedure, we exhaustively searched for negative samples. Detection results on INRIA dataset are more accurate compared to benchmark method HOG. We used monocular camera to estimate pedestrian distance which is fairly accurate. We apply our detector on real-time video without region of interest (ROI) selection and could achieve 7 fps detection speed.

The Electrocardiogram is a tool used to access the electrical recording and muscular function of the heart and in last few decades it is extensively used in the investigation and diagnosis of heart related diseases. It must be noted that the heart rate fluctuates not only because of cardiac demand, however is also influenced as a result of the occurrence of cardiac disease and diabetes. In addition, it has been shown that Heart Rate Variability (HRV) may well be utilized as an early indicator of cardiac disease susceptibility and the existence of diabetes. As a result, the HRV can be exercised for early clinical test of these diseases. Most existing systems make use of Support Vector Machine (SVM), owing to the generalization performance, it is not sufficient for the accurate classification of heart rate data. In order to overcome this complication, Improved Extreme Learning Machine (IELM) classifier is used, to obtain the best parameter value and best feature subset through the use of Bacterial Foraging Optimization (BFO) that feed the classifier. Here in this work, features of linear and nonlinear are extracted from the HRV signals. Following the preprocessing, feature extraction is done effectively together with feature selection with the assistance of BFO for the purpose of data reduction. Subsequently, proposed a scheme to integrate Kernel Fuzzy C-Means (KFCM) clustering and classifier to adequately enhance the accuracy result for ECG beat classification. The accuracy result for classification of heart rate data is shown in the proposed scheme.

Background/Objectives: Heart Rate Variability is an essential feature which decides the condition of human heart. ECG is used as diagnostic tool to access the electrical function of the heart. Methods/Statistical Analysis: The nine linear and nonlinear features are derived from the HRV signals. The feature extraction is carried out with the help of Particle Swarm Optimization (PSO) for data reduction. In proposed scheme Fuzzy C-Means (FCM) clustering and classifier integrated to enhance the accuracy result for ECG beat classification. Findings: The Enhanced SVM classifier classifies the heart rate data. Enhanced SVM classifier groups the linear and non-linear parameters as inputs, which are derived from the HRV signal. The denoise signals are classified and identifies the pattern for better classification of ECG signal. Application/Improvements: The proposed scheme is experimented with the assistance of the most commonly used MIT-BIH arrhythmia database and adequate results were obtained with an accuracy level of 98.38% than the other well-known approaches.

Due to the advancement in information and communication technology and vast varied learner group, e-learning has become popular. For achieving the adaptivity in learning, predefined concept map is used to provide proper guidance. In most of the researches, weight of concepts in each learning item is not properly considered. In this study, a three phase prerequisite concept map formulation for adaptivity is proposed. This approach is a most efficient one, since the first phase discards all the unrelated items which may distract the further analysis. Here, Norm-referencing in Item Analysis approach is used to find the item discrimination for elimination of irrelevant items. The second phase, computes all the grade association rules. The weight of concept in each learning item is considered and prerequisite concept sets are found which is not having any redundancy and cyclic in its map, thereby facilitating the next step of procedure. The final phase constructs the concept map with maximum confidence in a capable manner. Finally, the Prerequisite concept map can be used in tutoring system, thereby enhancing the adaptivity in e-learning.

Due to the advancement in information technology and varied learner group, e-learning has become popular. For achieving the adaptivity in learning, predefined concept map is used to provide proper guidance. Many researches propose various approaches in developing concept map by considering only binary numeric grade. Also, weight of concepts in each learning item is not properly considered. In this study, a Three-phase concept map design for adaptivity is proposed. This approach is an efficient one, since the first phase discards all unrelated items which may distract further analysis. Here, Norm-referencing in Item Analysis approach is used to find the item discrimination for elimination of irrelevant items. The second phase, computes all the grade association rules till Large_3 Item Sets. The weight of concept in each learning item is considered and prerequisite concepts are found which is not having any redundancy and cyclic in its map, thereby facilitating the next step of procedure. The final phase constructs the concept map with maximum confidence in a capable manner since most of the surplus computation is abolished in earlier phases. Finally, the concept map can be used in tutoring system, thereby enhancing the adaptivity in e-learning.

Computer based assessment is a prevalent method of administering the tests in the education environment. Randomization of test items here may produce unfair effect on test takers which is unproductive in the outcome of the test. Hence, the Intelligent Tutoring System came in to existence which assigns intelligent question depending on the student's response in the testing session. It will be more productive when the questions are assigned based on the learner ability in the early stage itself. So, this study focus on building up a framework to automatically assign intelligent question based on the learner ability not only all through the session, but also while entry. The learners are classified based on their level by Decision Tree using ID3 algorithm. At end, automatic up gradation of the question levels can be done dynamically. The questions are classified efficiently based on item difficulty using Item Response Theory. Item discrimination is found based on Point Biserial Correlation Coefficient and there by only the items which can discriminate the high and low performers alone are accumulated in the item pool to have maximum effect of intelligence in tutoring system. Thereby the objective of Intelligent Tutoring System is achieved by using both adaptability and intelligence in testing.

Due to the advancement in information technology and varied learner group, e-learning has become popular. Hence computer based assessment become a prevalent method of administering the tests. Randomization of test items here may produce unfair effect on test takers which is unproductive in the outcome of the test. There is a need to develop the Intelligent Tutoring System that assigns intelligent question depending on the student’s response in the testing session. It will be more productive when the questions are assigned based on the ability in the early stage itself. Also, if only the standard multiple-choice questions are focused, then the real embedded nature of computer assessment is sacrificed. Items with different constrained constructs are included to bring out the complex skills, analytical and comprehensive ability of learners. So, this study focus on building up a framework to automatically assign intelligent question with different constructs based on the learner ability while entry. Using Norm Referencing, questions are classified based on item difficulty. Item discrimination is found and there by only the items which can discriminate the performers alone are accumulated in the item pool to have maximum effect of intelligence in tutoring system. The level of new learner is predicted by means of Naive Bayesian classification and the consequent item is posed. Thereby the objective of Intelligent Tutoring System is achieved by using both adaptability and intelligence in testing.