Student Performance Prediction: Mukul Gharpure, Pushpak Chaudhari, Yash Bhole, Sagar Borkar, Aashutosh Awasthi

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 07 Issue: 08 | Aug 2020 www.irjet.net p-ISSN: 2395-0072
Student Performance Prediction

Mukul Gharpure1, Pushpak Chaudhari 2, Yash Bhole 3, Sagar Borkar 4, Aashutosh Awasthi5
1,2,3,4,5Department of Computer Engineering, Thakur College of Engineering & Technology, Mumbai.

---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - no and attendance were removed due to irrelevancy. After
In today’s World, life becomes easier with access to this we have calculated the correlation between the
information beforehand. This is where Data Science plays an attributes and after applying regression model, we found the
important role. We have prepared an algorithm where we values of the intercept and the corresponding estimated
can predict Student’s performance using various factors such error, significance code and the R-squared values i.e.
as his previous grades, his current academic performance
accuracy.
and then based on these factors we have visualize
performance using graphs. We have collected real time data
from each student and using these real time data we have Along with this we have visualized the data in the
applied algorithms on that and then we got our final result. histogram, correlation-gram indicating the number of
students in the particular range of pointer. Using the plot
In this project of ours, we have collected the data graph we have drawn the relation between the different
from students. We have asked them to provide us with their parameter of the statistics representing the accuracy of the
previous semester pointers, their current state Assignments,
regression.
Internal assessment marks, practical marks and end
semester marks. We have divided the data into two parts
which include 70 % of trained data and remaining as test 2. LITERATURE SURVEY: -
data. We have applied our model on trained data. We have a. NEED OF THE PROJECT:
used linear regression in our model.
Living in the world which is growing very fast with
First we will visualize our input data and then we digital methods and using old methods to do the work is not
will predict the range within which the grades of students a good way. Using Modern technology can reduce human
will fit. We have used correlation graph, correlation-gram
efforts and increase efficiency. In this way the smart work
and histogram in our visualization. And the visualisation will
be done with the help of colours for better understanding. process has helped us to implement this “Student
Performance Prediction” {for all Students & Teachers}. It is
Key Words: Linear Regression. designed for predicting the pointer which can be achieved in
coming semester. This application will help to find out the
1. PROBLEM DEFINITION: - performance of the student and if it is not good then we
student have to concentrate in improvement of the result.
We have collected data from every student of the This application is designed to implement data science and
class and thus forming the real time dataset. The dataset reduce manpower, time and human error in predicting the
taken contains the previous semester pointer and current same using no. of formulas. This system is also helpful for
semester pointer. We can segregate the data into trained and teacher to know about what type of improvement need in
test data. The 70% of the data is taken under train data. The student.
linear regression model is applied on this train data. By using
the Multivariate Linear Regression model we are able to find Communication gap between student and faculty is
out the relation between previous data and current pointer been reduced by this system such as students will always
and thus able to find out the equation between input and been informed about all the notices by manually, apply leave
output. and gate pass. Then for faculty apply leave and outdoor duty
applications.
Before applying the model the data is pre-processed
so as to remove non-uniformity. The row having an empty b. EXISTING SYSTEM:
attribute has been removed from the dataset which comes
under the data cleaning process. After data-processing, we The system of evaluating student in the college
have used only those attributes which were relevant to the nowadays is done using excel and applying formula. The
students’ performance prediction and columns such as roll formula calculation is very much complex and not that much
© 2020, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page 153
accurate. Also the visualization of that data is not done in DATA SET:
proper way. The number of rows and columns in excel sheet
is very large and not useful to find out something meaningful Ro Learn IAT Assign Practi Point
from that data. ll ing Aver ment cal er
3. ARCHITECTURE & BLOCK DIAGRAM: - No Attitu age Avg. Avg.
de
Avg.
1 94 11.5 97.33 93 8.72
2 72 0 45 42.5 5.56
3 80.5 11.5 93.33 85 8.76
4 86.5 12.5 85 85 8.4
FIGURE 1ARCHITECTURE AND BLOCK DIAGRAM Table 1 INPUT DATA SET
 COLLECTION OF DATA – Data is collected from the

CORRELATIONS:
provided excel file. We use the following data:
o Term Test marks average
o Previous semester pointer/percentage Figure 2 CORRELATION
o Learning Attitude
o Practical Marks
o Assignment marks
 CLEANING OF DATA – We pre-process the data to

remove some records with missing values. The
following methods are used to clean the dataset:
Figure 2 CORRELATION
o Tidyverse
o Filling missing value with zero It's a statistical measure that suggests the level of
linear dependence between two variables. If the correlation
 APPLYING A SUITABLE MODEL – We apply the coefficient is close to 1, it would indicate that the variables
correct algorithm model on the data. We specifically are positively linearly related and the scatter plot falls
use algorithms like Multi Variate Linear Regression. almost along a straight line with positive slope. For -1, it
indicates that the variables are negatively linearly related
 PERFORMANCE EVALUATION – Improving the and the scatter plot almost falls along a straight line with
accuracy of model with the help of true data. negative slope. And for zero, it would indicate a weak linear
relationship between the variables.
 RESULT INTERPRETATION – Use the models
developed and the graphs plotted make prediction. LINEAR REGRESSION:
Regression methods are used for estimating
numeric data. It explores the relationship between a
dependent, that is, the target variable and independent
variables or the predictor variables. This technique is used
for forecasting, time series modelling and finding the causal
effect relationship between the variables.
Linear regression can be used to find the

relationship between one predictor and target variable or
between target and more than one predictor variables.
Hence, linear regression is divided into 2 types:
 Simple Linear Regression y = b0 + b1*x1 + b2*x2 + b3*x3

 Multiple Linear Regression
The “b” values are called the regression weights or beta
Simple Linear Regression coefficients.
It establishes the relationship between two variables Example:

using a straight line. Hence, only one predictor is used.
Simple linear regression draws a line that comes closest to > input <- mtcars[,c("mpg","disp","hp","wt")]
the data by finding the slope and intercept that define the
line and minimize regression errors. > model <- lm(mpg~disp+hp+wt, data = input)
It searches for statistical relationship, not > summary(model)

deterministic relationship. Relationship between two
variables is known to be deterministic if one variable can be Here, mpg is the target variable and disp, hp and wt are the
accurately expressed by the other. For example, we can predictor variables.
easily predict the temperature in Fahrenheit if we already
the temperature in Celsius. Statistical relationship is not R-SQUARED:
accurate in determining relationship between two variables.
For example, the relationship between height and weight is R-squared is a statistical measure of how close the
not the most accurate in all cases. data are to the fitted regression line. It is also known as the
coefficient of determination. It is the percentage of the
The fundamental idea is to obtain a line that best fits response variable variation that is explained by a linear
the data. The best fit line is the one for which total prediction model.
error is as small as possible. Error is the distance between
the points to the regression line. Function used for building R-squared = Explained variation / Total variation
linear models is lm (). Where, R = 0<R<1
Example: 0% indicates that the model explains none of the

variability of the response data around its mean. 100%
> linearMod <- lm(dist ~ speed, data=cars) indicates that the model explains all the variability of the
response data around its mean. In general, the higher the R-
# build linear regression model on full data squared, the better the model fits your data.
> print(linearMod) Formula: R2 = 1 – (σ2/Var(Y))
Multiple Linear Regression ADJUSTED R-SQUARED:
Multiple linear regression attempts to model the Adjusted R-squared adjusts the statistic based on
relationship between two or more explanatory variables and the number of independent variables in the model. Adjusted
the response variable by fitting a linear equation to observed R2 also indicates how well terms fit a curve or line, but
data. Multiple regression is an extension of simple linear adjusts for the number of terms in a model. If you add more
regression. It is used when we want to predict the value of a and more useless variables to a model, adjusted r-squared
variable based on the value of two or more other variables. will decrease. If you add more useful variables, adjusted r-
The variable we want to predict is called the dependent squared will increase.
variable (or sometimes, the outcome, target or criterion
variable). It is also known as Multivariate regression. VISUALIZATIONS:
If we have 3 dependant variables, (x), the prediction With ever increasing volume of data, it is impossible
of y is expressed by the following equation:- to tell stories without visualizations. Data visualization is an
art of how to turn numbers into useful knowledge.
R Programming lets you learn this art by offering a

set of inbuilt functions and libraries to build visualizations
and present data. Before the technical implementations of
the visualization, let’s see first how to select the right chart
Figure 4 DATA.FRAME IN CORRGRAM
5. IMPLEMENTATION:-
STEP 1: ALGORITHM
A simple example of regression is predicting

Semester 3 pointer of a student when his previous
Figure 3 TYPES OF VISUALIZATION records are known. To do this we need to find the
relationship between his different academic year
performances.
type.
The steps to create the relationship is:
 Carry out the experiment of gathering a sample of

Selecting the Right Chart Type observed values of his previous semester pointers.
 Create a relationship model using the lm
There are four basic presentation types: () functions in R.
 Find the coefficients from the model created and
 Comparison create the mathematical equation using these
 Composition  Get a summary of the relationship model to know
the average error in prediction. Also
 Distribution called residuals.
 To predict the pointer of current Semester of new
 Relationship persons, use predict () function in R.
4. FLOWCHART:-
In this project, to arrive at a prediction, we use the
model of linear regression. In our dataset, we have
columns like the term test average, term work, practical
and assignments which help us determine a student’s
pointer. We also have the real time data of the pointers
that the students achieved so that we can train the
model accordingly and predict the pointers of students
in the test dataset.
relation. The predicted value is comparable to predicted

values. We have able to able to visualize the output even
with the help of histogram and correlation-gram. The
histogram shows the number of the students in the
particular pointer range and the correlation-gram is able to
show the accuracy. In this way the project has full-filled its
aim.
Figure 5 FLOWCHART
Figure 4 FLOWCHART
Figure 7 HISTOGRAM
Figure 6 DATA.FRAME IN CORRELATION
6. RESULT & DISCUSSION:-

After acquiring the real-time data for this project, we
have validated the data and then it is pre-processed to
remove the dis-continuity in the data. Thus, we are able to
derive a proper statistical relation between the various
attributes considered in the data like Term-test marks an
internal assessment marks, etc. After applying the
multivariate linear regression model on the trainee’s data set
we are able to get the equation between input parameter
and the output parameter i.e the predicted pointer. The
trainee’s data is obtained by splitting the dataset into train
and test data. Over 80% of the data comes under the train
data. The applying the machine learning model to trainee’s
data and putting the output of the model with the test data
set containing only the input parameter, we are able to
predict the end semester pointer. This pointer is then Figure 8 RESIDUAL OUTPUT
compared with the real pointer, and the R-squared value for
the project i.e. the coefficient of determination, is around
90%, which is pretty good. The R-squared value can further
be increased by increasing the number of rows in the dataset
so as to get even more accurate value of the intercepts in the
Figure 9 MODEL OUTPUT
Table 2 PREDICTED Vs REAL OUTPUT
Roll No Predicted Real

Pointers Pointers
14 8.5064501 8.84
38 7.8242863 7.33
50 8.1139504 8.67
51 9.2212334 8.72 Figure 10 PREDICTED VS REAL POINTERS
62 8.1893634 8.76
Prediction Accuracy: 0.8933519 i.e. 89%.
66 8.1422941 7.60
82 8.0497049 9.33 7. CONCLUSION:-
90 6.8556565 7.75
93 8.8362857 7.88 The project is pretty much successful. We are able to
94 8.1058232 9.24 understand the basics of the data science, get a quick insight
95 0.5693761 0.00 of this field and explore the different aspects of the same.
96 7.7706271 6.80 Over this period of time we have understood a few concepts
and have obtained good knowledge of how the project
should be done i.e. the critical steps that should be
considered while working on the project. The data cleaning
using suitable algorithm and different ways to represent the
data graphically so as to get an overview of the data
distribution and applying suitable machine learning model
to get desired result are some of the areas which we have
learned during the internship.
8. FUTURE SCOPE:-
Of course, there is so much to learn in this vast field. Also the

scope of improvisation is there in our project since we have
not be able to put it into a proper product. We will try to
make it more user-friendly by applying the user-interface
and increase the accuracy. We should be able to add even
more input attributes. In future, we will surely be able to
accomplish it by learning and clearing the concepts more
and off course with the help of our mentors of the internship.
Also with this project we can predict how the student will
perform in coming semester by using his previous
semesters’ pointers. In this way we can know which students
requires attention and could also prove helpful for the
college in the prediction of the passing percentage of the
students of the college and will be working on the same in
future.
9. REFERENCES:-
We have referred “http://www.kaggle.com/” for the data
set for practicing algorithms.
“https://www.tutorialspoint.com/r/r_multiple_regressio
n.htm” for learning our algorithm.
https://www.hackerearth.com/practice/machine-
learning/linear-regression/multivariate-linear-regression-
1/tutorial/
https://www.r-bloggers.com/r-tutorial-series-multiple-
linear-regression/
https://www.guru99.com/r-simple-multiple-linear-
regression.html

Student Performance Prediction: Mukul Gharpure, Pushpak Chaudhari, Yash Bhole, Sagar Borkar, Aashutosh Awasthi

Uploaded by

Copyright:

Available Formats

Student Performance Prediction: Mukul Gharpure, Pushpak Chaudhari, Yash Bhole, Sagar Borkar, Aashutosh Awasthi

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Student Performance Prediction: Mukul Gharpure, Pushpak Chaudhari, Yash Bhole, Sagar Borkar, Aashutosh Awasthi

Uploaded by

Copyright:

Available Formats

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056

Volume: 07 Issue: 08 | Aug 2020 www.irjet.net p-ISSN: 2395-0072

Student Performance Prediction

1,2,3,4,5Department of Computer Engineering, Thakur College of Engineering & Technology, Mumbai.

1 94 11.5 97.33 93 8.72

3 80.5 11.5 93.33 85 8.76

4 86.5 12.5 85 85 8.4

FIGURE 1ARCHITECTURE AND BLOCK DIAGRAM Table 1 INPUT DATA SET

 COLLECTION OF DATA – Data is collected from the

 CLEANING OF DATA – We pre-process the data to

Linear regression can be used to find the

 Simple Linear Regression y = b0 + b1*x1 + b2*x2 + b3*x3

It establishes the relationship between two variables Example:

It searches for statistical relationship, not > summary(model)

Example: 0% indicates that the model explains none of the

> print(linearMod) Formula: R2 = 1 – (σ2/Var(Y))

Multiple Linear Regression ADJUSTED R-SQUARED:

R Programming lets you learn this art by offering a

Figure 4 DATA.FRAME IN CORRGRAM

A simple example of regression is predicting

 Carry out the experiment of gathering a sample of

relation. The predicted value is comparable to predicted

Figure 6 DATA.FRAME IN CORRELATION

6. RESULT & DISCUSSION:-

Figure 9 MODEL OUTPUT

Table 2 PREDICTED Vs REAL OUTPUT

Roll No Predicted Real

Of course, there is so much to learn in this vast field. Also the

You might also like

 Simple Linear Regression y = b0 + b1x1 + b2x2 + b3*x3