Linear

Regression
Subjective
Questions
LIJIN L
 From your analysis of the categorical variables from the dataset, what could
you infer about their effect on the dependent variable?

I have plotted the categorical variables with the target variables on boxplot and has inferred
following effect on target.
Season: 3: fall has highest demand for rental bikes.
I see that demand for next year has grown.
Demand is continuously growing each month till June. September month has highest demand.
After September, demand is decreasing.
When there is a holiday, demand has decreased.
Weekday is not giving clear picture about demand.
The clear weathershit has highest demand.
Why is it important to use drop_first=True during dummy variable creation?

drop_first=True is important to use, as it helps in reducing the extra column

created during dummy variable creation. Hence it reduces the correlations
created among dummy variables.
If we do not drop one of the dummy variables created from a categorical
variable then it becomes redundant with dataset as we will have constant
variable(intercept) which will create multicollinearity issue.
Looking at the pair-plot among the numerical variables, which one has the
highest correlation with the target variable?

The feature “temp” has highest correlation. It is very well linearly

related with target “cnt”
How did you validate the assumptions of Linear Regression after building the
model on the training set?

Error terms are normally distributed with mean 0.

Error Terms do not follow any pattern.
Multicollinearity check using VIF(s).
Linearity Check.
Ensured the overfitting by looking the R2 value and Adjusted R2.
Based on the final model, which are the top 3 features contributing significantly
towards explaining the demand of the shared bikes?

Features “holiday”, “temp” and season “hum” are highly related with target
column, so these are top contributing features in model building.
 Explain the linear regression algorithm in detail.

Linear Regression Algorithm is a machine learning algorithm based on supervised learning. Linear regression is a part of
regression analysis. Regression analysis is a technique of predictive modelling that helps you to find out the
relationship between Input and the target variable.
Linear regression is one of the very basic forms of machine learning where we train a model to predict the behaviour of
your data based on some variables. In the case of linear regression as you can see the name suggests linear that means
the two variables which are on the x-axis and y-axis should be linearly correlated.
Example for that can be let’s say you are running a sales promotion and expecting a certain number of count of
customers to be increased now what you can do is you can look the previous promotions and plot if over on the chart
when you run it and then try to see whether there is an increment into the number of customers whenever you rate
the promotions and with the help of the previous historical data you try to figure it out or you try to estimate what will
be the count or what will be the estimated count for my current promotion this will give you an idea to do the planning
in a much better way about how many numbers of stalls maybe you need or how many increase number of employees
you need to serve the customer. Here the idea is to estimate the future value based on the historical data by learning
the behaviour or patterns from the historical data.
In some cases, the value will be linearly upward that means whenever X is increasing Y is also
increasing or vice versa that means they have a correlation or there will be a linear downward
relationship. One example for that could be that the police department is running a campaign
to reduce the number of robberies; in this case, the graph will be linearly downward.
Linear regression is used to predict a quantitative response Y from the predictor variable X.
Mathematically, we can write a simple linear regression equation as follow y ~ b0 + b1*x
Where y is the predicted variable (dependent variable), b1 is slope of the line, x is independent
variable, b0 is intercept(constant). It is cost function which helps to find the best possible value
for m and c which in turn provide the best fit line for the data points.
 Explain the Anscombe’s quartet in detail.

Anscombe's Quartet can be defined as a group of four data sets which are
nearly identical in simple descriptive statistics, but there are some peculiarities
in the dataset that fools the regression model if built. They have very different
distributions and appear differently when plotted on scatter plots. Each dataset
consists of eleven (x,y) points.
 What is Pearson’s R?

Pearson's r is a numerical summary of the strength of the linear association between the
variables. If the variables tend to go up and down together, the correlation coefficient will be
positive. Pearson's r measures the strength of the linear relationship between two variables.
Pearson's r always between -1 and 1. If data lie on a perfect straight line with negative slope,
then r = -1.
Positive correlation indicates the both the variable increase and decrease together. Negative
correlation indicates the one the variable increase and the other variable decrease and vice
versa.
What is scaling? Why is scaling performed? What is the difference between
normalized scaling and standardized scaling?

Scaling is a method to normalize the range of independent variables. It is performed to bring all
the independent variables on a same scale in regression. If Scaling is not done, then regression
algorithm will consider greater values as higher and smaller values as lower values.
It is important to note that scaling just affects the coefficients and none of the other
parameters like t-statistic, F-statistic, p-values, R-squared, etc.
Example Weight of a device = 500 grams, and weight of another device is 5 kg. In this example
machine learning algorithm will consider 500 as greater value which is not the case. And it will
do wrong prediction.
Machine Learning algorithm works on numbers not units. So, before regression on a dataset it
is a necessary step to perform. Scaling can be performed in two ways: Normalization: It scale a
variable in range 0 and 1. Standardization: It transforms data to have a mean of 0 and standard
deviation of 1
You might have observed that sometimes the value of VIF is infinite. Why does
this happen?

When there is a perfect relationship then VIF = Infinity whereas if all the
independent variables are orthogonal then to each other then VIF = 1.0. Means
if a variable is expressed exactly by a linear combination of other variable then
it is said that VIF is infinite.
What is a Q-Q plot? Explain the use and importance of a Q-Q plot in linear
regression.

Quantile-Quantile (Q-Q) plot, is a graphical tool to help us assess if a set of data plausibly came
from some theoretical distribution such as a Normal, exponential or Uniform distribution. Also,
it helps to determine if two data sets come from populations with a common distribution It is
used for determining if two data sets come from populations with a common distribution.
A q-q plot is a plot of the quantiles of the first data set against the quantiles of the second data
set. Whether the Distributions is Gaussian, Uniform, Exponential or even Pareto distribution, it
can be found out.
Few advantages: a) It can be used with sample sizes also
b) Many distributional aspects like shifts in location, shifts in scale, changes in symmetry, and
the presence of outliers can all be detected from this plot.
It is used to check following scenarios: If two data sets — i. come from populations with a
common distribution
ii. have common location and scale
iii. have similar distributional shapes iv. have similar tail behaviour
Different distribution: If all point of quantiles lies away from the straight line at an angle of 45
degree from x -axis