You are on page 1of 4

MACHINE LEARNING AND PATTERN

RECOGNITION (B9DA109_2223_TMD3)
CA_ONE_APRIL_2023.

Course Title: Master of Science in Data Analytics


Lecturer Name: Courtney ford
Module/Subject Title: Machine Learning and Pattern Recognition
Assignment Title: Supervised Machine Learning – Regression
Report Type: Individual
NAME: JAYA KRISHNA PEMMASANI

IDNO: 20000086
This is a continuous assignment report on supervise machine learning
and I have detailed explanation in the way how I was helped for
feature selection and which I have participated with our team
members to complete full assignment.
In this assignment I have involved in all the parts and follow up the
entire code of selecting input variables and valuables and adding rows
and columns and removing duplicate values etc… and mainly I have
involved in feature selection and model comparison.
A correlation matrix is created for understand and analyze the
correlation of variables among the input variables and against the
target variables.
In the code the input and output variable are not correlated then we
should drop them because input variable is not contributing for
outcome of prediction.
Generally feature selection is defined as a reducing the size of
datasets is important as the size and variety of datasets continue to
grow.

Here feature selection is used to the process more accurate and it


also useful to predict the values.
. For
scaling the process we have done a data transformation is used
here.

To reduce the over fitting in linear and logistic regression we use


one of the selection process is feature selection.
In feature selection we use some of the implementations they are
manually selecting important features and removing the un
necessary features and reducing noise and outliers in data set.
In my research I have learnt that feature selection has some types
techniques and methods they are supervised and un supervised
wrapper method and filter methods and Embedded methods and
these methods are each one divided into some sub parts etc..
Simply we can say that for developing a predictive method we can
use feature selection for reducing the input variables.

Learning’s and observations:


From this assignment I have learn and generating a heat map
with correlation matrix is to be better understand the relation
between variables.
Here we can get the results to be accuracy values from the given
data set based on a selection model.
In correlation matrix from the transformed data we have created
numerical values only.
We took a data set of fifteen variables dataset from that we have
given input data to some added rows and columns etc.

In model comparison we have used stochastic gradient Descent model.

By using SGD model we have found that we can predict an accurate


output values from the given input values and we can get the display
the data from the data set.
Conclusion:

Through this to predict the target values a base version model is


created and in linear regression a data set is created that contains
minimum and maximum values from that we can predict a target
values in a dataset

You might also like