Browsing by Author "Mohiuddin, Karishma"

Now showing 1 - 8 of 8

A Machine Learning Approach to Analyze and Reduce Features to a Significant Number for Employee’s Turn Over Prediction Model
(Scopus, 2020) Alam, Mirza Mohtashim; Mohiuddin, Karishma; Islam, Md. Kabirul; Hassan, Mehedi; Hoque, Md. Arshad-Ul; Allayear, Shaikh Muhammad
Turnover of employee considers as one of the major issue that every company faces. Especially, if the employee has advance skills at his/her working field, then the company faces great loss during that period. To find out the most dominant reasons of employee attrition, we approach by determining features and using machine learning algorithms where features have been processed and reduced beforehand. We have proposed a new model where particular attributes of employee turnover have been selected and adjusted accordingly. In first phase of our reduction method, Sequential Backward Selection Algorithm (SBS) has been used to reduce the features from a higher number to a relatively smaller significant number. After that Chi2 and Random Forest importance algorithm have been used together for the second phase of reduction to determine the common important features by both of the algorithms which can be considered as the foremost features that lead to employee turnover. Our two steps feature selection technique confirms that there are mainly three features that are responsible for employee’s departure. Later, these selected minimal features have been tested with state of the art algorithms of machine learning, such as Decision Tree, Random Forest, Support Vector Machine, Multi-layer Perceptron (MLP), K-Nearest Neighbor (kNN) and Gaussian Naïve Bayes. Lastly, the test result has been visualized by 3D representation to learn the features that are precisely involved for the employee’s turnover.
A Machine Learning Approach to Analyze and Reduce Features to a Significant Number for Employee’s Turn Over Prediction Model
(Springer Nature, 2018-11-02) Alam, Mirza Mohtashim; Mohiuddin, Karishma; Islam, Md. Kabirul; Hassan, Mehedi; Hoque, Md. Arshad-Ul; Allayear, Shaikh Muhammad
Turnover of employee considers as one of the major issue that every company faces. Especially, if the employee has advance skills at his/her working field, then the company faces great loss during that period. To find out the most dominant reasons of employee attrition, we approach by determining features and using machine learning algorithms where features have been processed and reduced beforehand. We have proposed a new model where particular attributes of employee turnover have been selected and adjusted accordingly. In first phase of our reduction method, Sequential Backward Selection Algorithm has been used to reduce the features from a higher number to a relatively smaller significant number. After that Chi2 and Random Forest importance algorithm have been used together for the second phase of reduction to determine the common important features by both of the algorithms which can be considered as the foremost features that lead to employee turnover. Our two steps feature selection technique confirms that there are mainly three features that are responsible for employee’s departure. Later, these selected minimal features have been tested with state of the art algorithms of machine learning, such as Decision Tree, Random Forest, Support Vector Machine, Multi-layer Perceptron (MLP), K-Nearest Neighbor and Gaussian Naïve Bayes. Lastly, the test result has been visualized by 3D representation to learn the features that are precisely involved for the employee’s turnover.
A Reduced feature based neural network approach to classify the category of students
(ACM, 2018-03-09) Alam, Mirza Mohtashim; Mohiuddin, Karishma; Das, Amit Kishor; Islam, Md Kabirul
To ensure more effectiveness in the learning process in educational institutions, categorization of students is a very interesting method to enhance student's learning capabilities by identifying the factors that affect their performance and use their categories to design targeted inventions for improving their quality. Many research works have been conducted on student performances, to improve their grades and to stop them from dropping out from school by using a data driven approach [1] [2]. In this paper, we have proposed a new model to categorize students into 3 categories to determine their learning capabilities and to help them to improve their studying techniques. We have chosen the state of the art of machine learning approach to classify student's nature of study by selecting prominent features of their activity in their academic field. We have chosen a data driven approach where key factors that determines the base of student and classify them into high, medium and low ranks. This process generates a system where we can clearly identify the crucial factors for which they are categorized. Manual construction of student labels is a difficult approach. Therefore, we have come up with a student categorization model on the basis of selected features which are determined by the preprocessing of Dataset and implementation of Random Forest Importance; Chi2 algorithm; and Artificial Neural Network algorithm. For the research we have used Python's Machine Learning libraries: Scikit-Learn [3]. For Deep Learning paradigm we have used Tensor-Flow, Keras. For data processing Pandas library and Matplotlib and Pyplot has been used for graph visualization purpose.
An Adaptive Feature Dimensionality Reduction Technique Based on Random Forest on Employee Turnover Prediction Model
(Springer Nature, 2018-10-26) Islam, Md. Kabirul; Alam, Mirza Mohtashim; Islam, Md. Baharul; Mohiuddin, Karishma; Das, Amit Kishor; Kaonain, Md. Shamsul
This paper is based on the theme of employee attrition where the reasoning behind employee turnover has predicted with the help of machine learning approach. As employee turnover has become a vital issue these days due to heavy work pressure, less salary, less work satisfaction, poor working environment; it’s high time to uphold a better solution on this term. Therefore, we have come up with a prediction model based on machine learning approach where we have used each feature’s respective Random Forest importance weights while threshold based correlated feature merging into each of the single combined variable. Again, we scale specific features to get the correlated matrix of features matrix by defining threshold. Certainly, this newly developed technique has achieved good result for some algorithms compared to Principal Component Analysis (PCA) and Linear Discriminant Analysis (LDA) for the same dataset.
Haar Cascade Classifier and Lucas–kanade Optical Flow Based Realtime Object Tracker with Custom Masking Technique
(Advances in Intelligent Systems and Computing, Springer, 2018-12-27) Mohiuddin, Karishma; Alam, Mirza Mohtashim; Das, Amit Kishor; Munna, Md. Tahsir Ahmed; Allayear, Shaikh Muhammad; Ali, Md. Haider
Computer vision has been proven a remarkable entity in modern computer science. Different applications of this field have been used on a regular basis. In this paper, we propose a new model for tracking real time object (i.e., car, human face, interior objects, arms, etc.) from video feed by providing training with HAAR features and also with the implementation of Lucas Kanade Optical Flow including Custom Masking technique. Object tracking has been considered to be very much useful in augmented reality, security, virtual reality, training with simulation, etc. In this research, we have trained our classifier of a specific object for detection purpose. Upon successful training of the classifier, a video footage has been passed into the classifier. Firstly, it detects region of interest (ROI) consisting of the object(s). Secondly, with necessary preprocessing techniques we have detected the contour area enclosing the object. Finding out the contour within the detected object(s) enabled us to create green color within the contour enclosing area. We only kept green contours and subtracted everything from the frames of the scene by Custom Masking technique. Finally, we tracked the object by trailing the green contours of the detected object(s) by Lucas Kanade Optical Flow. Our developed system is able to detect and track different object(s) from a video feed with the settings of less than or equal to 30 frames per second (FPS).
Haar Cascade Classifier and Lucas–Kanade Optical Flow Based Realtime Object Tracker with Custom Masking Technique
(Springer Nature, 2018-12-27) Mohiuddin, Karishma; Alam, Mirza Mohtashim; Das, Amit Kishor; Munna, Md. Tahsir Ahmed; Allayear, Shaikh Muhammad; Ali, Md. Haider
Computer vision has been proven a remarkable entity in modern computer science. Different applications of this field have been used on a regular basis. In this paper, we propose a new model for tracking real time object (i.e., car, human face, interior objects, arms, etc.) from video feed by providing training with HAAR features and also with the implementation of Lucas Kanade Optical Flow including Custom Masking technique. Object tracking has been considered to be very much useful in augmented reality, security, virtual reality, training with simulation, etc. In this research, we have trained our classifier of a specific object for detection purpose. Upon successful training of the classifier, a video footage has been passed into the classifier. Firstly, it detects region of interest (ROI) consisting of the object(s). Secondly, with necessary preprocessing techniques we have detected the contour area enclosing the object. Finding out the contour within the detected object(s) enabled us to create green color within the contour enclosing area. We only kept green contours and subtracted everything from the frames of the scene by Custom Masking technique. Finally, we tracked the object by trailing the green contours of the detected object(s) by Lucas Kanade Optical Flow. Our developed system is able to detect and track different object(s) from a video feed with the settings of less than or equal to 30 frames per second (FPS).
Image recognition by deep learning
(BRAC University, 8/21/2017) Mohiuddin, Karishma; Das, Amit Kishor; Obaid, Habiba Bint; Ali, Md. Haider; Uddin, Dr. Jia
Object recognition has become a crucial topic in the field of computer vision. Poor qualities of images unable bring out the desired object as per expectancy. Many models have proposed to recognize object from image. However, most of these approaches hardly achieve high accuracy and precision. It creates a major obstacle to get correctness of the research because of the lighting, illumination, image quality, noise, ethnicity and various angels of similar objects. Therefore, we have proposed a novel approach to detect any object by CNN method including HAAR Cascade classifier where we first detect the most prominent features from scene using Haar Feature Based Cascade Classifier that has been introduced by Paul Viola and Michael Jones. In the second phase, the classification has been used for Convolutional Neural Network to detect the object automatically with better accuracy and more efficiently. It can determine any object after proper training and dataset manipulation. Our proposed method for image recognition has achieved very good accuracy than our expectation.
Statistical Analysis and Identification of Important Factors of Liver Disease Using Machine Learning and Deep Learning Architecture
(ACM International Conference Proceeding Series, ACM Digital Library, 2019-03-15) Islam, Md. Kabirul; Alam, Mirza Mohtashim; Rony, Md Rashad Al Hasan; Mohiuddin, Karishma
One of the most prominent and metabolically the most active organ of human body is liver, whose formation is built upon subtle organic compounds and responsible for storing the energy of a living body. Damaging this organ could lead to liver failure and possibly imply to a gruesome death. An early detection of primary causes of liver failure could revoke the adverse effect of liver diseases. Several metabolic compounds of human body such as Bilirubin, Total Proteins, Albumin, Alkaline, Alamine, Asparatate and outer factors such as gender, age etc. are considered as vital elements to find possible liver disease patient. In this paper, we have attempted to analyze the causal factors behind liver disease statistically and bring out the most significant factors. Again, we have introduced a reduced model based on the significance of logistic regression analysis, where we successfully eliminated highly co-related variables to get rid of multi co-linearity. A comparison has been conducted on the basis of the performance of the various machine learning and deep learning algorithms.