AU2020101294A4 - Student’s physiological health behavioural prediction model using svm based machine learning algorithm - Google Patents

Abstract

STUDENT'S PHYSIOLOGICAL HEALTH BEHAVIOURAL PREDICTION MODEL USING SVM BASED MACHINE LEARNING ALGORITHM ABSTRACT Students behavior is especially important for educator in the education field. The faculties are responsible for smooth running of the class and at the same time, they should monitor the student's activities, etc. In recent trends, machine learning and computer vision technologies has become important role in education field. Machine learning is the concept of learning from the past and based on that improves the future performance. Machine learning algorithms are used to identify the student's behavior. Many challenges are there to face in students' performance. The aim is to find the students physiological health behavior that will help to improve their behavior and provides feedback to the teacher. The teachers can also capture the student's behavior in the classroom for the decision-making process and it is taken as a data-acquisition part. Computer vision technologies, machine learning algorithms and data analysis are used for monitoring task as well as student's behavior can be observed. Through the network, the student's behavior can be noted, and others can react if any critical situation is present. The machine learning model helps in healthcare model, monitoring student's health status using the parameters and classifies the student's health status using the algorithm Support Vector Machine. 11 P a g e STUDENT'S PHYSIOLOGICAL HEALTH BEHAVIOURAL PREDICTION MODEL USING SVM BASED MACHINE LEARNING ALGORITHM Diagram Data Retrievmg Face Recogntion and Detection Module Behavioural Module Face chme DMetec on Motig AnalymiaModulk Motion Hua Muttono Figure: Proposed Framework 1 P a g e

Description

STUDENT'S PHYSIOLOGICAL HEALTH BEHAVIOURAL PREDICTION MODEL USING SVM BASED MACHINE LEARNING ALGORITHM

Diagram

Data Retrievmg Face Recogntion and Detection Module Behavioural Module

Face chme

DMetec on

Motig AnalymiaModulk Motion Hua

Muttono

Figure: Proposed Framework

1 Pag e

STUDENT'S PHYSIOLOGICAL HEALTH BEHAVIOURAL PREDICTION MODEL USING SVM BASED MACHINE LEARNING ALGORITHM

Description

STUDENTS PHYSIOLOGICAL HEALTH BEHAVIOUR

Physical health is the most important one for learning process because illness of health disturbs the learning capacity. Physiological is the concept to identify the human nature, emotions, their thoughts, day of learning, memory power, sense perception, physical health, age, levels of fatigue, quality food and drink and atmospheric condition. The physiological factors are related to human physical body and it affects the thinking of the human. Some various physiological factors are discussed below: i) Sense and Perception: Sense perception is the concept of all basic cognitive learning. If the perception power is less, the learning will become lesser. A blind person learning will be lesser than the normal person. Physical or mental impairment also lowers the learning process. ii) Physical Health: If the person has physical health illness, the learning will become lesser. If the physical health is good, it helps to pursue the more learning activities for a long-term education or longer education. An unhealthy person will be affected in normal physical strength which is needed for any of the mental activities. iii) Fatigue: Fatigue such as muscular or sensor may lead to laziness and boredom. In school or home environment, the factors may cause mental and fatigue. The mental and physical fatigue is due to seating arrangement is bad, lacking accommodation, unhealthy dressing, no proper ventilation, poor light, noise in over crowd, and nutrition. Learning capacity may also get affected due to fatigue because of longer time study in home. iv) Time of learning: To concentrate on studies, morning and evening is the best time to study. During the day time, there will be less mental capacity. The more variation will be there in day time to learn efficiently. The Time of learning helps to know how far the new knowledge is acquired.

v) Food and Drink: For efficient mental activity, nutrition is the most important factor. Poor nutrition and some of the food also affects the learning capacity for the humans. vi) Atmospheric condition: Mental efficiency may also decrease due to variation in temperature and humidity level. Learning capacity may also get disturbed due to improper ventilation, poor illumination, noise in over crowd schools, physically not comfortable in over crowd school. Various diversions and distractions may affect the concentration in studying and learning. If people are in difficult environment, they will struggle to learn the new knowledge. vii) Age: Age also matters in learning capacity. Some of the subjects can be known at earlier age and some of the subjects may known in adulthood age. Mental development will increase up to the age of 23 and it halts at the age of 40. Learning process will be rapid between the age of 18 and 20, maintains stagnant level at the age of 25, and starts to decrease up to 35. Mental maturation will be determined by the age factor. Some of the critical problems cannot be solved by the person until he gets matured sufficiently. For example: While comparing with uneducated adults, the children will learn the school subjects in fast manner. Because, their nervous system will be flexible and they are not burdened with complex worldly problems.

DECISION TREE

Decision tree (DT) is the concept of constructing a tree from the input data. A class or value will be determined, so that the tree can extract a set of rules. Sequential data are used by decision trees in order to make separate groups and maximize the distance between each group. In knowledge based decision tree classification algorithms, the output is constructed in a tree structure from different states of adjective values. The main advantage is its super flexibility and comprehensibility. DT is an effective tool for predicting different categories. The DT is a successful method and it is used as knowledge-based expert systems.

SUPPORT VECTOR MACHINE

Support vector machine concept is to search for a hyperplane in an N-dimensional space and N is the number of features. The hyperplane will classify the data points. To form a different classis from data points, a different variety of hyperplanes can be chosen. The aim of the SVM algorithm is to find out the best boundary between the data. It considers the possible distance from all different categories and will not be sensitive to other data points. SVM is classification algorithm and it is considered as the best technique to categorize and predict, detecting the data. SVM is a supervised learning category and has two phases of training and testing.

YOLO- YOUONLYLOOKONCE

The Main difference between target detection algorithms and classification algorithms is that, in detection algorithms, a bounding box can be provided to surround the object and that will be located within the image. R-CNN will take more amount of time to train the network. As it takes over time for each test image, so it cannot be implemented in real-time. The selective search algorithm is a fixed algorithm and learning doesn't take place here. The alternative algorithm is Fast R-CNN, which is significantly quicker in training and testing sessions comparative to R-CNN. Both algorithms R-CNN & Fast R-CNN uses selective search algorithm to find out the region proposals. The Selective Search is a slow and time-consuming process that affects the performance of the network. Therefore, the object detection algorithm is introduced to eliminate the selective search algorithm. This system allows to learn the region proposals. In Faster R-CNN, the image is provided as an input to a convolutional network which provides a convolutional feature map. To identify the region proposals, a separate system is used to predict the region proposal. Using the Rol pooling layer, the expected region proposals are reshaped. So, this is used to classify the image within the proposed region, also predicts the offset values for the bounding boxes. Faster R-CNN is much faster than R-CNN and Fast R-CNN. Therefore, it can be used for real-time object detection.

The previous object detection algorithms use regions to localize the object within the image, but the network does not concentrate on the complete picture. YOLO or You Only Look Once is an object detection algorithm and it is very much different from the region-based algorithms, which is discussed above. In YOLO, a single convolutional network predicts the bounding boxes and the class probabilities for the tables. Consider an input image and split it into an S x S grid, 1. consider m bounding boxes within each of the grid, 2. the network results a class probability as well as offsets values for each of the bounding boxes, 3. In bounding boxes, if the class possibility is above a threshold value, that is selected and used to locate the object within the image. YOLO is faster than other object detection algorithms. The only limitation of YOLO algorithm is that it struggles with small objects within the image, for example, Difficulty in detecting a flock of birds and this is due to the spatial constraints of the algorithm.

BLOCKDIAGRAM

The aim is to find the students physiological health behavior that will help to improve their behavior and also provides feedback to the teacher. The objective is that, the teachers can also capture the student's behavior in the classroom for the decision-making process and it is taken as a data-acquisition part. The details are recorded to identify the student's attention in the class. Due to some challenges, students may get affected in the study life. Students achievement will be based on the faculties, educational programs, environmental conditions, study hours, academic details and performance, financial problems, and institutional environment. If the student's behavior is bad, the teacher can grasp the behavior of student and they can make some reasonable points to adjust in order to change the learning environment for the students. A teacher can identify the good and bad behavior of the student in the real environment by observing and questioning them. All these are possible for small number of students, but it is not possible for large number of students. The effective tools are needed for tracking the large number of students which is not possible in real environment. Without any human efforts, the database can be collected to know the student behavior accurately. In research analysis, the researchers have found that, the students who are playing or texting to some one during the lecture class which is taken on digital device results in affecting the long-term memory. In this way, the students will get poor marks in the exams even though, if they do not affect by the short-term memory. If the student's concentration is more in class, they will have higher chance for achievement. If their mind works in multi-task, the brain will lose some one's attention. With the help of computer vision applications, student's behavior can be recorded and analyzed. By using this data, the faculty can also note the negative attitude of the student during class time and they can know the reasons. In existing system, the student's behavior is guessed based on their studying performance. Computer vision technologies helps to monitor the student's performance during the lecture. The main focus can be done on the eye directions, head movement, facial expression and the movement of the body. In eye directions, we can notice whether the student focusing on teacher or board, notebook or bench, and other directions. In computer vision concept, the camera which is placed in classroom is used as a data collector module. The camera is connected to processing unit for the purpose of data analysis, data storage. By using this, behavioral analysis can be done based on facial-recognition, body movement. The body detection is categorized as face detection, upper body detection and full body detection. The factors can be noted in body detection such as seating and concentrating in classroom, seating but student didn't concentrate in the class room, standing and leaving the classroom, yawning, leaning back, eye-movement all these attitudes are shown in classroom by the students. The student's monitoring will be directly observed by the camera to know the details. The camera records the video and the data will be retrieved. This data will be fed to frame processing to output the facial expressions, eye movement, and head movement, body movement. The data will be analyzed and summarized for future process. After summarizing the data, the facial component is used in order to classify. Face detection method is used to detect the face of the student. Face detection is the part of the object detection. Machine learning is the concept of learning from the past experience and based on that increases the performance. It helps to analyze the data from the past database and predicts the future performance. The machine learning algorithms helps for object detection. The object detection algorithm is preferred; the data is given as input to a Deep Learning YOLO to realize the objects out of different classes. Based on the data present, proper operations can be planned and executed. Convolutional neural networks (CNNs) are one of the solutions to recognize objects on an image. YOLO can recognize objects on images. Feature extraction will be helpful for face recognition. Using face embedding algorithms, the student face can be recognized. Students eyes should also be recognized but the eyes will not be clear in blurry image, etc. The positions can also be recognized, for example: two students can view similar directions, but the objects which is observed may be different. Face reaction, position movement, behavioral observation, body movements are the main factors to observe and analyze the student's concentration level. During class time, attendance and student's behavior must be noted simultaneously by the faculty. It is a complex job and results in affecting the both teaching and learning. The data analysis module is integrated with computer vision technologies and the machine learning algorithms in order to understand the student's behavior. Therefore, the processing unit is divided into modules such as face recognition, motion analysis, and behavioral analysis. Face recognition helps to identify the students face for the identification process. So that, their behavioral analysis can be known. In motion analysis, their activities can be noted. The student's upper body and full body can be detected. Image Pre-processing is a necessary base in most of the image processing concepts. In this phase, pre-processing consists of Image De-noising, and image enhancement. Image de-noising can be done using the filtering concept to reduce some disturbance present in an image. Image enhancement can be done using algorithms to improve the contrast and brightness of the image. Image pre-processing is done before the segmentation process in order to remove some imperfections like a reduction of bugs in images, reduction of noise, non-uniform illumination, lighting variations, to remove reflections, masking portions of images. Filtering is the most important concept to improve the image quality and to clear the noise present in the images. Image blurring can be done to remove some background variations. Image smoothing is used to improve the clarity of the image. In the original image, an uneven lighting distribution may be present due to uneven illumination conditions during imaging. To overcome the uneven illumination, brightness normalization is needed to apply to the image. Later, the RGB image will be converted to Gray-scale image. In motion-analysis, the parameters to be find out is the movement of the body. In body detection, the upper body and full body is to be detected. The students will be either sitting or standing in the classroom, for this purpose the body movements has to be noted. The images are given as input to the machine learning algorithms to classify the output. The machine learning model helps in healthcare model, monitoring student's health status using the parameters and classifies the student's health status using SVM.

Claims (5)

STUDENT'S PHYSIOLOGICAL HEALTH BEHAVIOURAL PREDICTION MODEL USING SVM BASED MACHINE LEARNING ALGORITHM We claim,

1. Computer vision technologies and machine learning algorithms has become significant in education filed. Computer vision technologies helps to monitor the student's behavior during the lecture class, their facial expression, motion analysis, and behavioral analysis.

2. Machine learning algorithms will be helpful in classifying the output as student's health status, behavior analysis, motion analysis, facial recognition, and gaze directions.

3. Support Vector Machine is a classification algorithm, it helps to classify and predicts the data. The pre-processing algorithms used to eliminate all the noise content in the image.

4. The Enhancement technique is used to improve image brightness and contrast.

5. YOLO algorithm is used to detect the object in an image and works faster. Without any human efforts, the database can be collected to know the student behavior accurately.

1 Pag e

STUDENT’S PHYSIOLOGICAL HEALTH BEHAVIOURAL PREDICTION MODEL 09 Jul 2020

USING SVM BASED MACHINE LEARNING ALGORITHM

Diagram 2020101294

Figure: Proposed Framework

1|Page