2.

Discuss on the given performance measures: MAE, MSE, RMSE, RSE, R2 score in
the context of Linear Regression. Justify the usability of each measure.

n the context of linear regression, which aims to model the relationship between
independent variables and a continuous dependent variable, various performance
measures are used to assess the quality of the model's predictions. Let's discuss
each of these measures and their usability:
Mean Absolute Error (MAE):
MAE measures the average absolute difference between the predicted values and
the actual values.
It provides a straightforward interpretation as it gives the average magnitude of
errors in the predictions.
Usability: MAE is useful when you want to understand the average error
magnitude without considering the direction of errors. It's less sensitive to outliers
compared to other measures like MSE.
Mean Squared Error (MSE):
MSE measures the average squared difference between the predicted values and
the actual values.
Squaring the errors penalizes larger errors more heavily, making MSE sensitive to
outliers.
Usability: MSE is widely used as it's easy to compute and differentiable. However,
it's sensitive to outliers, which can skew the evaluation of the model's
performance.
Root Mean Squared Error (RMSE):
RMSE is the square root of MSE, which brings the error metric back to the same
scale as the dependent variable.
It provides an interpretable measure of the average magnitude of errors, similar
to MAE but with sensitivity to outliers due to the squaring.
Usability: RMSE is useful for providing a measure of the typical deviation of the
predictions from the actual values, taking into account the scale of the dependent
variable. It's commonly used and easy to interpret.
Residual Standard Error (RSE):
RSE measures the standard deviation of the residuals (the differences between
observed and predicted values).
It's an absolute measure of lack of fit, providing a measure of the typical size of
the residuals.
Usability: RSE is useful for assessing the goodness of fit of the model. Lower values
indicate better fit, and it's particularly useful for comparing different models.
Coefficient of Determination (R-squared or R2 Score):
R2 score measures the proportion of the variance in the dependent variable that
is predictable from the independent variables.
It ranges from 0 to 1, where 1 indicates a perfect fit.
Usability: R2 score is widely used to evaluate the goodness of fit of the model. It
provides an intuitive interpretation of the proportion of variance explained by the
model. However, it can be misleading if used alone, especially when dealing with
complex models or multicollinearity.
In summary, each of these performance measures has its own strengths and
weaknesses, and their usability depends on the specific context and requirements
of the problem at hand. It's often recommended to use a combination of these
measures to gain a comprehensive understanding of the model's performance.

What is the need for feature scaling? Differentiate standard scaling and
minmax scaling.
Feature scaling is a crucial preprocessing step in many machine learning
algorithms, including linear regression, support vector machines (SVM), and
k-nearest neighbors (KNN). It involves transforming the features of the
dataset to a similar scale, which helps in improving the performance and
convergence of the machine learning algorithms. The need for feature
scaling arises due to the following reasons:
Different Scales: Features in a dataset often have different scales. For
example, one feature may range from 0 to 100 while another ranges from
1000 to 10000. These differences in scale can lead to biased or skewed
results in algorithms that rely on distance calculations or gradient descent
optimization.
Gradient Descent Optimization: Algorithms like linear regression and neural
networks use optimization techniques like gradient descent to minimize a
cost function. Features with larger scales can dominate the optimization
process, causing it to take longer to converge or to converge to suboptimal
solutions.
Distance-Based Algorithms: Algorithms like KNN and SVM use distance
metrics to make predictions. If features are not scaled, features with larger
scales will have a higher impact on the distance calculations, leading to
biased predictions.
Now, let's differentiate between two common methods of feature scaling:
Standard Scaling (Z-score normalization) and Min-Max Scaling.
Standard Scaling (Z-score normalization):
In standard scaling, each feature is scaled to have a mean of 0 and a
standard deviation of 1.
The formula for standard scaling is: 𝑧=𝑥−𝜇𝜎z=σx−μ
This method centers the data around 0 and scales it to have a standard
deviation of 1.
Standard scaling is useful when the distribution of the features is
approximately Gaussian (normal distribution).
Min-Max Scaling:
In min-max scaling, each feature is scaled to a fixed range, usually between 0
and 1.
The formula for min-max scaling is: 𝑥scaled=𝑥−min(𝑥)max(𝑥)
−min(𝑥)xscaled=max(x)−min(x)x−min(x)
This method scales the data to a specific range, making it suitable for
algorithms that require features to be on the same scale.
Min-max scaling preserves the original distribution of the data and is less
affected by outliers compared to standard scaling.
In summary, both standard scaling and min-max scaling are used to scale
features to a similar range, reducing the impact of feature scale differences
on the performance of machine learning algorithms. Standard scaling is
suitable for normally distributed data, while min-max scaling is useful when
preserving the original distribution of the data is important. The choice
between these methods depends on the characteristics of the dataset and
the requirements of the algorithm being used.

Top of Form
Q…What is Decision Tree? Briefly discuss on the terminologies used in
Decision Tree

Decision tree is a type of supervised learning algorithm (having a predefined

target variable) that is mostly used in classification problems. It can also be
used for solving regression problem. ●It works for both categorical and
continuous input and output variables. ●In this technique, we split the
population or sample into two or more homogeneous sets (or sub-
populations) based on most significant splitter differentiator in input
variables.
Discuss the decision tree splitting criteria Gini and Entropy