PMPaul P. Mathai

Paul P. Mathai

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Paul P. Mathai has a total of 15 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Machine Learning and Deep Learning Applications in Groundwater Quality Prediction: A Comprehensive Survey

    by Aiswarya Vijayakumar, Paul P Mathai

    Groundwater quality is critical for drinking, irrigation, and ecosystem health. Recent studies have leveraged machine learning (ML) and deep learning (DL) to model and predict groundwater quality parameters and indices more accurately and efficiently than traditional methods. This survey reviews significant research from the past few years on applying ML/DL techniques to groundwater quality analysis. We summarize each study's objectives, study area and data, modeling techniques, key input variables, performance metrics, and findings. Common approaches include predicting water quality indices (WQI) or specific contaminants algorithms including ensemble approaches, decision trees, artificial neural networks (ANN), support vector machines (SVM), and recurrent neural networks. According to the survey, ML/DL models offer insights into the most important water quality indicators and typically achieve excellent forecast accuracy for groundwater quality. A comparative summary of studies is presented to highlight methodological differences and performance outcomes. We wrap up with insights into prevailing challenges, such as data heterogeneity and model generalizability, and outlines future research directions.


  • Open Access

    Prediction of Right Bowlers for Death Overs in Cricket

    by Lakshmi Shaju, Rahul K Sajith, N P Reshma, N Vinayak, Paul P Mathai

    Predicting the right bowlers for the death overs in cricket is crucial for the success of a team, as these overs can often determine the outcome of a match. Death overs refer to the final overs of an innings, which are usually considered to be the most crucial because they can determine the result of the match. The death overs are typically the last 5-6 overs of an innings, and are known for being high-pressure situations due to the need to score runs or take wickets. Batting team has always an advantage in winning the match, by selecting the right bowlers to bowl in death overs, it would create an edge for the bowling team to win the match. Selecting the right bowlers for death overs requires an in-depth understanding of the strengths and weaknesses of different bowlers and the ability to analyze various factors such as the pitch conditions, the opposition team, and the current game situation. Some of the key factors that can be used to predict the right bowlers for death overs include the bowler’s past performance in similar situations, their ability to bowl under pressure, and their ability to take wickets, their ability to bowl Yorkers and slower balls. In order to accurately predict the right bowlers for death overs, it is important to consider all of these factors and use statistical analysis and machine learning techniques to build models that can accurately predict their performance. This analysis can then be used to build models that can accurately predict which bowlers are likely to perform well in the death overs based on these factors.


  • Envisaging the Academic Performance of Students

    by Ierin Babu, R V Siva Balan, Paul P Mathai

    Certain Colleges students’ performance will be improved during their studies can be planned is important fortelling the studies performance of students. Here using the performance evaluation is measured on the basis of higher secondary level and, analysis of Emotional Quotient and Intelligent Quotient. An evaluation is doing according to the evaluation that is received from the analysis based on th mentioned above. The emotional quotient and intelligent quotient is not always same, it be far different in various people being. But these two content will be helping in the development of their academic knowledge and life long learning. In this paper presenting the performance of students by their higher secondary value, IQ and EQ values. It is find out only through some novel algorithms. It is defined on the basis of training, learning, analysis and prediction stage.Firstly students or faculty have to enter their marks in the interface provided and students have to attend an intelleigent quotient and emotional quotient questions that is set with the instruction of some psychologist. In prediction stage will producewheather the students will pass or fail. The testing phase it helps to determine is our prediction is correct or not.


  • An Efficient Data Mining Algorithm Based on Item Sets Frequency and Priority Using Distribution Model

    by Paul P Mathai, R. V. Siva Balan

    Frequent pattern mining plays an essential role in many important data mining tasks, e.g., associations, sequential patterns, partial periodicity, etc. But, it is also well known that frequent pattern mining often generates a very large number of frequent item sets and rules, which reduces not only efficiency but also effectiveness of mining since users have to sift through a large number of mined rules to find useful ones. So instead of mining the frequent itemsets, mining only closed frequent itemsets will effectively increase the efficiency and also reduces the computation time. In our proposed closed frequent item set mining, the data are collected from the database. After gathering all the data, support of each itemsets are computed by moving the sliding window. Following that, the frequent itemsets are mined by defining a predefined threshold. The obtained frequent itemsets are then given to the Genetic Algorithm (GA) for the optimization purpose. A hash table is maintained to preserve the closed frequent itemsets. Subsequently, the consistency based on the probability distribution function followed by utility is computed for the mined closed frequent itemsets and the itemsets which having the high distribution value are selected as the closed frequent itemsets. The proposed closed frequent itemset mining technique is implemented in the MATLAB platform and the performance of the proposed technique is compared with the existing techniques.


  • Automatic 2D to 3D video and image conversion based on global depth map

    by Shelmy Mathai, Paul P Mathai, K A Divya

    3d technology brings a new era of entertainment to the human race. It offered a wide array of possibilities in near future in almost every walk of life and in entertainment segment. 3D content generation is the important step in 3D systems. Special cameras such as stereoscopic dual camera, depth range camera etc. are designed to generate the 3D model of a scene directly. There are different techniques to generate the 3D content. But the problem is our current and past media data are in 2D which needs to convert into 3D. This is where the importance of 2D to 3D transformation arises. In this paper proposed real time 3D image and video creation by depth map estimation. Depth map estimation can be done in two methods. One is based on depth fusion method and other is based on saliency map of an image. In dataset image estimate the depth map from depth fusion method and then depth is refined by color spatial variance. In non dataset images we find depth map by global saliency method. Experimental result demonstrates that the proposed technique convey better performance compared to the state-of-the-art of methods.


  • An extensive review of facial expression recognition using salient facial patches

    by Elsa Sabu, Paul P Mathai

    Derivation of discriminant features from outstanding face patches has a major role in accurate identification of face expressions. The precise discovery of fragments in the face image improves the confinement of the outstanding patches on face images. Some methods uses a special framework for expression identification by the use of appearance related features of important face patches. A few noticeable face patches which are alive for each expression classes are extracted considering the location of fragments in the face image. Then further process these alive patches in order to obtain the patches which are discriminant and contain features for grouping of each pair of expressions. Thus it selects different facial patches as important for each pairs of expression classes. Then use these features for performing one against one categorization method. Also, an automated learning free method for fragment identification can be adopted, which achieves similar performances as that of other related fragment detection methods, and requires less execution time compared to others.

  • Salient Region Detection Based on Spatial Weight Map

    by Shelmy Mathai, Paul P. Mathai

    Saliency detection is defined to be a key attention mechanism by enabling organisms to focus their limited perceptual and cognitive resources of available sensory data. In short, saliency detection is nothing but detecting the more attracted regions in an image. Like a loud noise in a quite environment saliency is the contras-tic difference between the visually attracted items and their neighborhood. Detecting those attracted areas in the image is termed to be saliency detection. We proposed a novel salient region detection method by integrating four features namely boundary information, frequency weight map, global contrast and color spatial variance. Finally, the saliency map is defined as being the average of the three feature maps added by color spatial variance. Experimental result shows that the proposed method produce better performance compared to the state-of-the-art of methods. The programming and simulation of the processes as well as the analysis of the results were done using MATLAB.