6/28/23, 4:20 PM Regression Metrics ( Prudhvi Vardhan Notes) - Jupyter Notebook

Regression Metrics
Regression metrics are used to evaluate the performance of a regression model. They
measure the difference between the predicted and actual values, and can be used to compare
different models or to track the performance of a model over time.

We cannot calculate accuracy for a regression model.

The skill or performance of a regression model must be reported as an error in those

predictions.

This makes sense if you think about it. If you are predicting a numeric value like a height or
a dollar amount, you don’t want to know if the model predicted the value exactly (this might
be intractably difficult in practice); instead, we want to know how close the predictions were
to the expected values.

1. MAE( Mean Absolute Error)

2. MSE(Mean Squared Error)

3. RMSE ( Root Mean Squared Error)

4. R2 ( R squared)

5. Adj R2 ( Adjusted R squared)

localhost:8888/notebooks/ Regression Metrics ( Prudhvi Vardhan Notes).ipynb 1/26

6/28/23, 4:20 PM Regression Metrics ( Prudhvi Vardhan Notes) - Jupyter Notebook

Here are some detailed about regression metrics:

Mean Absolute Error (MAE): This metric measures the average of the absolute
differences between the predicted and actual values. It gives equal weight to all errors and
is less sensitive to outliers.

Mean Squared Error (MSE): This metric measures the average of the squared differences
between the predicted and actual values. It is more sensitive to outliers than MAE.

Root Mean Squared Error (RMSE): This metric is the square root of MSE. It is in the
same units as the response variable, making it easier to interpret.

R-squared (R2): This metric measures the proportion of variance in the target variable
explained by the model. It ranges from 0 to 1, with higher values indicating better
performance.

Adjusted R-squared (Adj R2): This metric is similar to R2 but accounts for the number of
predictors used in the model. It penalizes overfitting by adjusting for the number of
predictors.

Which metric to use?

The choice of metric depends on the specific problem and the desired outcome.

MAE : is a good choice when the goal is to minimize the overall error in the model while
avoiding large errors.

RMSE: is a good choice when the goal is to measure the absolute fit of the model.

R-squared: is a good choice when the goal is to explain the variability in the target
variable using the predictors.

Adj R2 : is a good choice when the goal is to select a parsimonious model that explains
the variability in the target variable.

Here is a table summarizing the different metrics:

Metric Description Units Interpretation

Mean Absolute Same as the Average of the absolute differences between the
MAE
Error response variable predicted and actual values

Same as the
Mean Squared Average of the squared differences between the
MSE response variable
Error predicted and actual values
squared

Square root of the average of the squared

Root Mean Same as the
RMSE differences between the predicted and actual
Squared Error response variable
values

Coefficient of Proportion of variance in the target variable

R-squared 0 to 1
Determination explained by the model

localhost:8888/notebooks/ Regression Metrics ( Prudhvi Vardhan Notes).ipynb 2/26

6/28/23, 4:20 PM Regression Metrics ( Prudhvi Vardhan Notes) - Jupyter Notebook

Metric Description Units Interpretation

1.MAE ( Mean Absolute Error )

Mean Absolute Error (MAE) is a regression metric that measures the average of the absolute
differences between the predicted and actual values. It is a robust metric, meaning that it is not
as sensitive to outliers as other metrics, such as Mean Squared Error (MSE).

Formula:

MAE = 1/n * Σ|y_true - y_pred|

y_true: The actual value of the target variable.

y_pred: The predicted value of the target variable.

n: The number of observations.

Interpretation:

The MAE is the average distance between the predicted and actual values. A lower MAE
indicates that the model is closer to the actual values, while a higher MAE indicates that
the model is further away from the actual values.

localhost:8888/notebooks/ Regression Metrics ( Prudhvi Vardhan Notes).ipynb 3/26

6/28/23, 4:20 PM Regression Metrics ( Prudhvi Vardhan Notes) - Jupyter Notebook

Here are some of the advantages and disadvantages of MAE:

Advantages:

Robust to outliers: MAE is not as sensitive to outliers as other metrics, such as MSE.
This means that it is less likely to be affected by a few extreme values in the data.

Easy to interpret: MAE is easy to understand and interpret. It is simply the average of the
absolute differences between the predicted and actual values.

Scale-independent: MAE is scale-independent, which means that it is not affected by the

units of the data. This makes it a good choice for comparing models that are trained on
different datasets with different scales.

Disadvantages:

Not as sensitive to large errors: MAE is not as sensitive to large errors as other metrics,
such as MSE. This means that it may not be as good at detecting models that are
consistently over- or under-estimating the target variable.

Not differentiable: MAE is not differentiable, which means that it cannot be used with
gradient descent optimization algorithms.

Overall, MAE is a good choice for evaluating the performance of a regression model when the
goal is to minimize the overall error in the model while avoiding large errors. However, it is
important to be aware of the limitations of MAE, such as its lack of sensitivity to large errors.
localhost:8888/notebooks/ Regression Metrics ( Prudhvi Vardhan Notes).ipynb 4/26
6/28/23, 4:20 PM Regression Metrics ( Prudhvi Vardhan Notes) - Jupyter Notebook

Here is a table summarizing the advantages and disadvantages of MAE:

Advantage Disadvantage

Robust to outliers Not as sensitive to large errors

Easy to interpret Not differentiable

Scale-independent

Main goal of MAE is Reduce the error - outcome should be Low or small , & if you have more
outliers then choose MAE

2.MSE ( Mean Squared Error)

Mean Squared Error (MSE) is a regression metric that measures the average of the squared
differences between the predicted and actual values. It is a sensitive metric, meaning that it is
more sensitive to outliers than other metrics, such as MAE.

its also used as Loss Function

Formula:

MSE = 1/n * Σ(y_true - y_pred)^2

y_true: The actual value of the target variable.

y_pred: The predicted value of the target variable.
n: The number of observations.

Interpretation:

localhost:8888/notebooks/ Regression Metrics ( Prudhvi Vardhan Notes).ipynb 5/26

6/28/23, 4:20 PM Regression Metrics ( Prudhvi Vardhan Notes) - Jupyter Notebook

The MSE is the average squared distance between the predicted and actual values. A lower
MSE indicates that the model is closer to the actual values, while a higher MSE indicates that
the model is further away from the actual values.

Image:

Example:

Let's say we have a regression model that predicts house prices. The actual house prices are
100,000𝑡𝑜
in the range of 500,000. If the MSE for the model is
100,000,𝑡ℎ𝑒𝑛𝑡ℎ𝑖𝑠𝑚𝑒𝑎𝑛𝑠𝑡ℎ𝑎𝑡𝑡ℎ𝑒𝑚𝑜𝑑𝑒𝑙𝑖𝑠,𝑜𝑛𝑎𝑣𝑒𝑟𝑎𝑔𝑒, 100,000 away from the actual house
prices, squared.

How to choose MSE:

The MSE metric is a good choice when the goal is to minimize the overall error in the model,
even if it means that some outliers are penalized more heavily. For example, if the target
variable is a continuous variable with a narrow range of values, then the MSE metric may be a
good choice.

Other metrics:

There are other regression metrics that can be used to evaluate the performance of a model,
such as MAE and RMSE. The choice of metric depends on the specific problem and the
desired outcome
localhost:8888/notebooks/ Regression Metrics ( Prudhvi Vardhan Notes).ipynb 6/26
6/28/23, 4:20 PM Regression Metrics ( Prudhvi Vardhan Notes) - Jupyter Notebook

Advantages and disadvantages of MSE:

Advantages:

Sensitive to large errors: MSE is more sensitive to large errors than other metrics, such
as MAE. This means that it is better at detecting models that are consistently over- or
under-estimating the target variable.

Differentiable: MSE is differentiable, which means that it can be used with gradient
descent optimization algorithms.

Disadvantages:

Not robust to outliers: MSE is not as robust to outliers as other metrics, such as MAE.
This means that it may be affected by a few extreme values in the data.

Not scale-independent: MSE is not scale-independent, which means that it is affected by

the units of the data. This makes it a less good choice for comparing models that are
trained on different datasets with different scales.

Overall, MSE is a good choice for evaluating the performance of a regression model when the
goal is to minimize the overall error in the model, even if it means that some outliers are
penalized more heavily. However, it is important to be aware of the limitations of MSE, such as
its lack of robustness to outliers.

Here is a table summarizing the advantages and disadvantages of MSE:

Advantage Disadvantage

Sensitive to large errors Not robust to outliers

Differentiable Not scale-independent

RMSE ( Root Mean Squared Error)

Root Mean Squared Error (RMSE) is a regression metric that measures the average
magnitude of the errors between the predicted and actual values. It is a sensitive metric,
meaning that it is affected by outliers.

Formula:

RMSE = √(1/n * Σ(y_true - y_pred)^2)

y_true: The actual value of the target variable.

y_pred: The predicted value of the target variable.

n: The number of observations.

localhost:8888/notebooks/ Regression Metrics ( Prudhvi Vardhan Notes).ipynb 7/26

6/28/23, 4:20 PM Regression Metrics ( Prudhvi Vardhan Notes) - Jupyter Notebook

Interpretation:

The RMSE is the square root of the average squared error. A lower RMSE indicates that the
model is closer to the actual values, while a higher RMSE indicates that the model is further
away from the actual values.

Image:

Example:

Let's say we have a regression model that predicts house prices. The actual house prices are
100,000𝑡𝑜
in the range of 500,000. If the RMSE for the model is
50,000,𝑡ℎ𝑒𝑛𝑡ℎ𝑖𝑠𝑚𝑒𝑎𝑛𝑠𝑡ℎ𝑎𝑡𝑡ℎ𝑒𝑚𝑜𝑑𝑒𝑙𝑖𝑠,𝑜𝑛𝑎𝑣𝑒𝑟𝑎𝑔𝑒, 50,000 away from the actual house
prices, but the errors are not as sensitive to outliers as they would be with MAE.

How to choose RMSE:

The RMSE metric is a good choice when the goal is to minimize the absolute fit of the model.
For example, if the target variable is a continuous variable with a wide range of values, then the
RMSE metric may be a good choice.

localhost:8888/notebooks/ Regression Metrics ( Prudhvi Vardhan Notes).ipynb 8/26

6/28/23, 4:20 PM Regression Metrics ( Prudhvi Vardhan Notes) - Jupyter Notebook

Other metrics:

There are other regression metrics that can be used to evaluate the performance of a model,
such as MAE and R-squared. The choice of metric depends on the specific problem and the
desired outcome.

Advantages of RMSE:

Sensitive to large errors: RMSE is more sensitive to large errors than other metrics, such
as MAE. This means that it is better at detecting models that are consistently over- or
under-estimating the target variable.

Scale-independent: RMSE is scale-independent, which means that it is not affected by

the units of the data. This makes it a good choice for comparing models that are trained on
different datasets with different scales.

Disadvantages of RMSE:

Not as robust to outliers: RMSE is not as robust to outliers as other metrics, such as
MAE. This means that it may be affected by a few extreme values in the data.

Not as easy to interpret: RMSE is not as easy to understand and interpret as other
metrics, such as MAE. It is the square root of the average squared error, which can be
difficult to understand for non-technical audiences.

Overall, RMSE is a good choice for evaluating the performance of a regression model when the
goal is to minimize the absolute fit of the model. However, it is important to be aware of the
limitations of RMSE, such as its sensitivity to outliers and its difficulty to interpret.

Here is a table summarizing the advantages and disadvantages of RMSE:

Advantage Disadvantage

Sensitive to large errors Not as robust to outliers

Scale-independent Not as easy to interpret

R2 (r_Squared)

R-squared (R2) is a regression metric that measures the proportion of variance in the target
variable explained by the model. It is a normalized metric, meaning that it is not affected by the
units of the data.

Formula:

R^2 = 1 - (SSR/SST)

SSR: The sum of squared residuals.

SST: The total sum of squares.

localhost:8888/notebooks/ Regression Metrics ( Prudhvi Vardhan Notes).ipynb 9/26

6/28/23, 4:20 PM Regression Metrics ( Prudhvi Vardhan Notes) - Jupyter Notebook

Interpretation:

R2 is a percentage, and it ranges from 0 to 1. A higher R2 indicates that the model is better at
explaining the variance in the target variable. A perfect model would have an R2 of 1, which
means that the model perfectly explains the variance in the target variable.

Advantages:

Easy to understand and interpret: R2 is easy to understand and interpret. It is simply the
proportion of variance in the target variable that is explained by the model.

Normalized: R2 is normalized, which means that it is not affected by the units of the data.
This makes it a good choice for comparing models that are trained on different datasets
with different scales.

Disadvantages:

Sensitive to outliers: R2 is sensitive to outliers, meaning that a few extreme values in the
data can significantly affect the value of R2.

localhost:8888/notebooks/ Regression Metrics ( Prudhvi Vardhan Notes).ipynb 10/26

6/28/23, 4:20 PM Regression Metrics ( Prudhvi Vardhan Notes) - Jupyter Notebook

Not a good measure of absolute fit: R2 is not a good measure of absolute fit, meaning
that it does not tell you how close the predicted values are to the actual values.

Overall, R2 is a good choice for evaluating the performance of a regression model when the
goal is to measure the proportion of variance in the target variable that is explained by the
model. However, it is important to be aware of the limitations of R2, such as its sensitivity to
outliers.

Here is a table summarizing the advantages and disadvantages of R2:

Advantage Disadvantage

Easy to understand and interpret Sensitive to outliers

Normalized Not a good measure of absolute fit

Adjusted R_Square

Adjusted R-squared is a metric that measures the proportion of variance in the target variable
explained by the model, adjusted for the number of predictors used in the model. It is a
modification of the R-squared metric that penalizes the model for adding additional predictors
that do not significantly improve the fit of the model.

Formula:

Adj. R2 = {1 - (1-R^2)(n-1) / (n-k-1)}

n: the number of data points in our dataset

k: the number of independent variables

R: The R-squared values determined by the model.

Interpretation:

The Adj. R2 is a measure of how well the model fits the data, taking into account the number of
predictors in the model. A higher Adj. R2 indicates that the model fits the data better.

localhost:8888/notebooks/ Regression Metrics ( Prudhvi Vardhan Notes).ipynb 11/26

6/28/23, 4:20 PM Regression Metrics ( Prudhvi Vardhan Notes) - Jupyter Notebook

Advantages:

Adjusts for the number of predictors: Adj. R2 penalizes the model for adding additional
predictors that do not significantly improve the fit of the model. This makes it a more
reliable measure of the model's performance than R-squared, which can be inflated by
adding irrelevant predictors.

Easy to interpret: Adj. R2 is easy to interpret. It is simply a percentage that indicates the
proportion of variance in the target variable that is explained by the model.

Disadvantages:

Sensitive to outliers: Adj. R2 is sensitive to outliers, meaning that it can be affected by a

few extreme values in the data.

Not scale-invariant: Adj. R2 is not scale-invariant, meaning that it is affected by the units
of the data. This makes it a less reliable measure of the model's performance when the
data is measured in different units.

Overall, Adj. R2 is a good choice for evaluating the performance of a regression model when
the goal is to measure the proportion of variance in the target variable that is explained by the
model, taking into account the number of predictors in the model. However, it is important to be
aware of the limitations of Adj. R2, such as its sensitivity to outliers and its lack of scale
invariance.

Here is a table summarizing the advantages and disadvantages of Adj. R2:

localhost:8888/notebooks/ Regression Metrics ( Prudhvi Vardhan Notes).ipynb 12/26

6/28/23, 4:20 PM Regression Metrics ( Prudhvi Vardhan Notes) - Jupyter Notebook

Advantage Disadvantage

Adjusts for the number of predictors Sensitive to outliers

Easy to interpret Not scale-invariant

Difference Between R-squared. and Adjusted R-

squared.

R-squared is a metric that measures the proportion of variance in the target variable
explained by the model. It is a measure of how well the model fits the data. R-squared
ranges from 0 to 1, with 1 indicating a perfect fit and 0 indicating no fit.

Adjusted R-squared is a metric that measures the proportion of variance in the target
variable explained by the model, adjusted for the number of predictors used in the model. It
is a modification of the R-squared metric that penalizes the model for adding additional
predictors that do not significantly improve the fit of the model.

The main difference between R-squared and Adjusted R-squared is that Adjusted R-
squared takes into account the number of predictors in the model. This means that
Adjusted R-squared is a more reliable measure of the model's performance than R-
squared, which can be inflated by adding irrelevant predictors.

For example, if you have a model with 2 predictors and an R-squared of 0.8, this means
that the model explains 80% of the variance in the target variable. However, if you add a
third predictor that does not significantly improve the fit of the model, the R-squared may
increase to 0.85. However, the Adjusted R-squared will not increase as much, because it
takes into account the fact that you have added an additional predictor.

In general, Adjusted R-squared is a better measure of the model's performance than R-

squared, especially when the model has a large number of predictors. However, R-squared is
still a useful metric, and it can be used to compare models with different numbers of predictors.

Here is a table summarizing the difference between R-squared and Adjusted R-squared:

Metric Description

Measures the proportion of variance in the target

R-squared
variable explained by the model.

localhost:8888/notebooks/ Regression Metrics ( Prudhvi Vardhan Notes).ipynb 13/26

6/28/23, 4:20 PM Regression Metrics ( Prudhvi Vardhan Notes) - Jupyter Notebook

Metric Description

Measures the proportion of variance in the target

Adjusted R-squared variable explained by the model, adjusted for the
number of predictors used in the model.

Penalizes the model for adding additional

predictors that do not significantly improve the fit of No
the model.

More reliable measure of the model's performance

Not as reliable as Adjusted R-squared but still a useful

The key difference between R-squared and Adjusted R-squared is that Adjusted R-squared
accounts for the complexity of the model by considering the number of predictors. It
provides a more accurate measure of the model's goodness of fit and helps in comparing
models with different numbers of predictors.

In summary, R-squared measures the proportion of variance explained by the

independent variables without considering the model's complexity, while Adjusted R-
squared adjusts for the number of predictors, providing a more reliable measure of the
model's explanatory power.

In [7]: ## Code
​
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
​
df =pd.read_csv("D:\\datascience\\Nitish sir\\Machine Learning\\placement.csv"
​
df

Out[7]:
cgpa package

0 6.89 3.26

1 5.12 1.98

2 7.82 3.25

3 7.42 3.67

4 6.94 3.57

... ... ...

195 6.93 2.46

196 5.89 2.57

197 7.21 3.24

198 7.63 3.96

199 6.22 2.33

200 rows × 2 columns

localhost:8888/notebooks/ Regression Metrics ( Prudhvi Vardhan Notes).ipynb 14/26

6/28/23, 4:20 PM Regression Metrics ( Prudhvi Vardhan Notes) - Jupyter Notebook

In [8]: df.shape

Out[8]: (200, 2)

In [9]: ​
# Plot a scatter plot of CGPA vs Package using the provided DataFrame.
plt.scatter(df['cgpa'],df['package'])
​
plt.xlabel('CGPA')
plt.ylabel('Package(in lpa)')

Out[9]: Text(0, 0.5, 'Package(in lpa)')

In [10]: # Extracting the first column of the DataFrame into X

X = df.iloc[:, 0:1]
​
# Extracting the last column of the DataFrame into y
y = df.iloc[:, -1]

localhost:8888/notebooks/ Regression Metrics ( Prudhvi Vardhan Notes).ipynb 15/26

6/28/23, 4:20 PM Regression Metrics ( Prudhvi Vardhan Notes) - Jupyter Notebook

In [12]: X

Out[12]:
cgpa

0 6.89

1 5.12

2 7.82

3 7.42

4 6.94

... ...

195 6.93

196 5.89

197 7.21

198 7.63

199 6.22

200 rows × 1 columns

In [13]: y

Out[13]: 0 3.26
1 1.98
2 3.25
3 3.67
4 3.57
...
195 2.46
196 2.57
197 3.24
198 3.96
199 2.33
Name: package, Length: 200, dtype: float64

In [14]: from sklearn.model_selection import train_test_split

​
# Splitting the data into training and testing sets
X_train,X_test,y_train,y_test = train_test_split(X,y,test_size=0.2,random_stat

localhost:8888/notebooks/ Regression Metrics ( Prudhvi Vardhan Notes).ipynb 16/26

6/28/23, 4:20 PM Regression Metrics ( Prudhvi Vardhan Notes) - Jupyter Notebook

In [15]: from sklearn.linear_model import LinearRegression

​
# Create an instance of LinearRegression
lr = LinearRegression()
​
# Fit the model using the training data
lr.fit(X_train, y_train)

Out[15]: ▾ LinearRegression
LinearRegression()

In [16]: # Plot a scatter plot of 'cgpa' versus 'package' and overlay the linear regres
plt.scatter(df['cgpa'],df['package'])
​
#line
plt.plot(X_train,lr.predict(X_train),color='red')
​
#labels
plt.xlabel('CGPA')
plt.ylabel('Package(in lpa)')

Out[16]: Text(0, 0.5, 'Package(in lpa)')

In [17]: from sklearn.metrics import mean_absolute_error,mean_squared_error,r2_score

​

localhost:8888/notebooks/ Regression Metrics ( Prudhvi Vardhan Notes).ipynb 17/26

6/28/23, 4:20 PM Regression Metrics ( Prudhvi Vardhan Notes) - Jupyter Notebook

In [18]: # Predict using the linear regression model on the test data
lr.predict(X_test)

Out[18]: array([3.89111601, 3.09324469, 2.38464568, 2.57434935, 1.6537286 ,

1.77647803, 2.07219258, 2.93143862, 3.76278706, 2.93701814,
4.09197872, 3.51170867, 2.97049525, 2.40138424, 3.18809652,
3.46707251, 1.94386362, 3.24389172, 2.97607477, 3.41685683,
2.55761079, 3.16577844, 2.85890486, 3.12114229, 3.68467378,
2.8700639 , 3.49497011, 3.34432308, 3.91901361, 1.96060218,
3.65119666, 3.2104146 , 3.74046898, 2.7863711 , 2.78079158,
3.27178932, 3.52844723, 2.61340599, 2.65804215, 2.71383735])

In [19]: y_pred =lr.predict(X_test)

In [20]: #Actual Values

y_test.values

Out[20]: array([4.1 , 3.49, 2.08, 2.33, 1.94, 1.48, 1.86, 3.09, 4.21, 2.87, 3.65,
4. , 2.89, 2.6 , 2.99, 3.25, 1.86, 3.67, 2.37, 3.42, 2.48, 3.65,
2.6 , 2.83, 4.08, 2.56, 3.58, 3.81, 4.09, 2.01, 3.63, 2.92, 3.51,
1.94, 2.21, 3.34, 3.34, 3.23, 2.01, 2.61])

In [21]: # MAE
​
print("MAE",mean_absolute_error(y_test,y_pred))
​
​
# Result - lpa Units

MAE 0.2884710931878175

Explanation:
It calculates the Mean Absolute Error (MAE) between the predicted values (y_pred) and the
actual values (y_test) in a regression problem. The MAE value is then printed.

mean_absolute_error is a function or method that calculates the MAE between two sets
of values.

y_test represents the actual or true values of the target variable in the regression problem.

y_pred represents the predicted values of the target variable obtained from the regression
model.

The "mean_absolute_error" function computes the absolute difference between each predicted
value and its corresponding actual value. It then calculates the average of these absolute
differences to obtain the MAE.

The output of the code would be:

MAE 0.2884710931878175

localhost:8888/notebooks/ Regression Metrics ( Prudhvi Vardhan Notes).ipynb 18/26

6/28/23, 4:20 PM Regression Metrics ( Prudhvi Vardhan Notes) - Jupyter Notebook

"MAE" is a label indicating the metric being printed.

The value 0.2884710931878175 is the computed MAE between the predicted and actual
values. It represents the average absolute difference between the predicted and actual values
in the regression problem. The lower the MAE value, the better the model's performance,
as it indicates smaller prediction errors on average.

Overall, this code snippet allows you to calculate and print the MAE value as a measure of the
model's performance in terms of the absolute difference between predicted and actual values.

In [24]: # MSE
​
print("MSE",mean_squared_error(y_test,y_pred))
​
​
# Result - Not in lpa Units

MSE 0.12129235313495527

Explanation :
It calculates the Mean Squared Error (MSE) between the predicted values (y_pred) and the
actual values (y_test) in a regression problem. The MSE value is then printed.

mean_squared_error is a function or method that calculates the MSE between two sets of
values.

y_test represents the actual or true values of the target variable in the regression problem.

y_pred represents the predicted values of the target variable obtained from the regression
model.

The mean_squared_error function computes the squared difference between each predicted
value and its corresponding actual value. It then calculates the average of these squared
differences to obtain the MSE.

The output of the code would be:

MSE 0.12129235313495527

"MSE" is a label indicating the metric being printed.

The value 0.12129235313495527 is the computed MSE between the predicted and actual
values. It represents the average squared difference between the predicted and actual
values in the regression problem. The lower the MSE value, the better the model's
performance, as it indicates smaller prediction errors on average.

Overall, this code snippet allows you to calculate and print the MSE value as a measure of the
model's performance in terms of the squared difference between predicted and actual values.

localhost:8888/notebooks/ Regression Metrics ( Prudhvi Vardhan Notes).ipynb 19/26

6/28/23, 4:20 PM Regression Metrics ( Prudhvi Vardhan Notes) - Jupyter Notebook

In [25]: # RMSE
​
print("RMSE",np.sqrt(mean_squared_error(y_test,y_pred)))
​
​
# Results - in Lpa Units

RMSE 0.34827051717731616

Explanation :
It calculates the Root Mean Squared Error (RMSE) between the predicted values (y_pred) and
the actual values (y_test) in a regression problem. The RMSE value is then printed.

mean_squared_error is a function or method that calculates the MSE between two sets of
values.

y_test represents the actual or true values of the target variable in the regression problem.

y_pred represents the predicted values of the target variable obtained from the regression
model.

np.sqrt is a function from the NumPy library that calculates the square root of a given
value.

The mean_squared_error function computes the squared difference between each predicted
value and its corresponding actual value. It then calculates the average of these squared
differences to obtain the MSE. Finally, np.sqrt is used to take the square root of the MSE,
resulting in the RMSE.

The output of the code would be:

RMSE 0.34827051717731616

"RMSE" is a label indicating the metric being printed.

The value 0.34827051717731616 is the computed RMSE between the predicted and
actual values. It represents the square root of the average squared difference between the
predicted and actual values in the regression problem. The lower the RMSE value, the
better the model's performance, as it indicates smaller prediction errors on average.

Overall, this code snippet allows you to calculate and print the RMSE value as a measure of
the model's performance in terms of the root of the squared difference between predicted and
actual values.

In [27]: # r2_Score
​
print("r2_Score:",r2_score(y_test,y_pred))
​

r2_Score: 0.780730147510384

localhost:8888/notebooks/ Regression Metrics ( Prudhvi Vardhan Notes).ipynb 20/26

6/28/23, 4:20 PM Regression Metrics ( Prudhvi Vardhan Notes) - Jupyter Notebook

Explanation:
it calculates the R-squared (R²) between the predicted values (y_pred) and the actual values
(y_test) in a regression problem. The R² value is then printed.

r2_score is a function or method that calculates the R-squared between two sets of
values.
y_test represents the actual or true values of the target variable in the regression problem.

y_pred represents the predicted values of the target variable obtained from the regression
model.

The r2_score function computes the R-squared, which measures the proportion of the total
variance in the dependent variable that can be explained by the independent variables in the
regression model.

r2: 0.780730147510384

"r2:" is a label indicating the metric being printed.

The value 0.780730147510384 is the computed R² between the predicted and actual
values. It represents the proportion of the variance in the dependent variable explained by
the independent variables in the regression model. R² ranges from 0 to 1, where 1
indicates that the model explains all the variability in the data.

Overall, this code snippet allows you to calculate and print the R² value as a measure of the
model's performance in explaining the variance in the dependent variable. A higher R² value
indicates a better fit of the model to the data.

In [28]: # r2
​
r2 = r2_score(y_test,y_pred)
​

In [29]: # Adjusted r2
​
X_test.shape
​

Out[29]: (40, 1)

In [30]: # Formula
​
1 - ((1-r2)*(40-1)/(40-1-1))
​

Out[30]: 0.7749598882343415

localhost:8888/notebooks/ Regression Metrics ( Prudhvi Vardhan Notes).ipynb 21/26

6/28/23, 4:20 PM Regression Metrics ( Prudhvi Vardhan Notes) - Jupyter Notebook

Explanation:
the provided shape of the test data, it seems that you are calculating the Adjusted R-squared
score manually using the formula:

Adjusted R² = 1 - ((1 - R²) * (n - 1) / (n - p - 1))

where:

R² is the R-squared score.

n is the number of samples (40 in this case).

p is the number of predictors or independent variables (1 in this case).

Let's break down the calculation using the provided values:

X_test.shape
(40, 1)

The shape of X_test indicates that you have 40 samples and 1 predictor variable.

1 - ((1 - r2) * (40 - 1) / (40 - 1 - 1))

0.7749598882343415

r2 refers to the R-squared score, which is 0.780730147510384 as calculated earlier.

Plugging in the values into the formula, we get 0.7749598882343415 as the calculated
Adjusted R² score.

The Adjusted R-squared score is used to account for the number of predictors in the model,
providing a penalty for adding unnecessary variables that do not contribute significantly
to the model's performance. In this case, the calculated Adjusted R² score suggests that the
model explains approximately 77.5% of the variance in the dependent variable, taking into
account the single predictor variable and the number of samples.

Importance of Adjusted R-Squared

localhost:8888/notebooks/ Regression Metrics ( Prudhvi Vardhan Notes).ipynb 22/26

6/28/23, 4:20 PM Regression Metrics ( Prudhvi Vardhan Notes) - Jupyter Notebook

In [31]: # Create a copy of the original DataFrame

new_df1 = df.copy()
​
# Add a new column 'random_feature' with random values between 0 and 1
new_df1['random_feature'] = np.random.random(200)
​
# Reorder the columns of the DataFrame
new_df1 = new_df1[['cgpa', 'random_feature', 'package']]
​
# Display the first few rows of the modified DataFrame
new_df1.head()

Out[31]:
cgpa random_feature package

0 6.89 0.533412 3.26

1 5.12 0.137277 1.98

2 7.82 0.529477 3.25

3 7.42 0.789529 3.67

4 6.94 0.467272 3.57

In [32]: # Random data is not linear

​
plt.scatter(new_df1['random_feature'],new_df1['package'])
plt.xlabel('random_feature')
plt.ylabel('Package(in lpa)')

Out[32]: Text(0, 0.5, 'Package(in lpa)')

localhost:8888/notebooks/ Regression Metrics ( Prudhvi Vardhan Notes).ipynb 23/26

6/28/23, 4:20 PM Regression Metrics ( Prudhvi Vardhan Notes) - Jupyter Notebook

In [45]: # Import necessary libraries

​
X = new_df1.iloc[:, 0:2]
y = new_df1.iloc[:, -1]
​
# Split the data into training and testing sets
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, rando
​
# Create and train the linear regression model
random_df = LinearRegression()
random_df.fit(X_train, y_train)
​
# Make predictions on the test set
y_pred = random_df.predict(X_test)
​
# Calculate and print the R2 score
print("R2 score:", r2_score(y_test, y_pred))
​
# Calculate the R2 score for further use
random_r2 = r2_score(y_test, y_pred)

R2 score: 0.7670139044847327

In [40]: # Here we can clearly see that R2 score has slightly decreased from
​
#r2_Score: 0.780730147510384 (Old) - random_r2_score: 0.7670139044847327 (rand
​

In [47]: # Adjusted R2_score

​
1 - ((1-random_r2)*(40-1)/(40-1-2))
​

Out[47]: 0.7544200614839075

In [ ]: # After apply Random data , Adjusted r2 has slightly decreased from

​
# Adjusted r2_score : 0.77495988823434159 (Old) - 0.7544200614839075 (random_d

Adding new Column With Linear Relationship

In [41]: new_df2 = df.copy()

​
new_df2['iq'] = new_df2['package'] + (np.random.randint(-12,12,200)/10)
​
new_df2 = new_df2[['cgpa','iq','package']]

localhost:8888/notebooks/ Regression Metrics ( Prudhvi Vardhan Notes).ipynb 24/26

6/28/23, 4:20 PM Regression Metrics ( Prudhvi Vardhan Notes) - Jupyter Notebook

In [42]: new_df2.head()

Out[42]:
cgpa iq package

0 6.89 3.26 3.26

1 5.12 2.08 1.98

2 7.82 2.25 3.25

3 7.42 3.37 3.67

4 6.94 2.77 3.57

In [43]: # Scatter plot of 'iq' vs 'package'

plt.scatter(new_df2['iq'],new_df2['package'])
plt.xlabel('iq')
plt.ylabel('Package(in lpa)')

Out[43]: Text(0, 0.5, 'Package(in lpa)')

localhost:8888/notebooks/ Regression Metrics ( Prudhvi Vardhan Notes).ipynb 25/26

6/28/23, 4:20 PM Regression Metrics ( Prudhvi Vardhan Notes) - Jupyter Notebook

In [52]: # Import necessary libraries

​
X = new_df1.iloc[:, 0:2]
y = new_df1.iloc[:, -1]
​
# Split the data into training and testing sets
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, rando
​
# Create and train the linear regression model
useful_df = LinearRegression()
useful_df.fit(X_train, y_train)
​
# Make predictions on the test set
y_pred = useful_df.predict(X_test)
​
# Calculate and print the R2 score
print("R2 score:", r2_score(y_test, y_pred))
​
# Calculate the R2 score for further use
useful_r2 = r2_score(y_test, y_pred)

R2 score: 0.7670139044847327

In [53]: 1 - ((1-useful_r2)*(40-1)/(40-1-2))

Out[53]: 0.7544200614839075

In [ ]: ​

localhost:8888/notebooks/ Regression Metrics ( Prudhvi Vardhan Notes).ipynb 26/26