Lan Gauge

Download as pptx, pdf, or txt
Download as pptx, pdf, or txt
You are on page 1of 17

Machine

Learning?

• Machine Learning
• Study of algorithms that improve their performance
at some task with experience
• Optimize a performance criterion using example data or
experience.
• Role of Statistics: Inference from a sample
• Role of Computer science: Efficient algorithms to
• Solve the optimization problem
• Representing and evaluating the model for inference 1
Machine learning is a field which
focuses on the use of data and
algorithms to imitate the way that
humans learn, gradually improving its
accuracy.

Machine learning is a field of study


that looks at using computational
algorithms to turn empirical data into
usable models.
• Features are the fields used as input.

Features
• A feature is one column of the data in your
input set.

• For instance, if you're trying to predict the


type of pet someone will choose, your input
features might include age, home region, family
income, etc.

• Feature means property of your training data.

• A feature is the input you have fed to the


model or system.

• The value of x variable in simple linear


regression are the features.
Label
• The output you get from your model
after training is called a label.

• A label is the thing we're predicting.

• For example the value of y variable in


simple linear regression model is the
label.

• Suppose you give your model data like


a person’s age, height, hair length and
then your model predicts whether the
person is male or female. Then male or
female is called the label.
Tabular data to predict whether car is stolen or not?

Features Label
• A model is the relationship between features and the
Model label.

• An ML model is a mathematical model that generates


predictions by finding patterns in your data.

• ML Models generate predictions using the patterns


extracted from the input data.

• A model represents what was learned by a machine


learning algorithm.

• The model is the “thing” that is saved after running


a machine learning algorithm on training data and
represents the rules, numbers, and any other
algorithm-specific data structures required to make
predictions.

ML
Data ML Model
Algorithm
Steps in ML
7.
Make
6. predicti
5. Parameter ons
4. Train Evaluate tuning
the the model
3. Choose
a ML model
2. Data
1. Data preparati model
collecti on
on
Approaches to Modelling
Techniques
in ML
The main machine learning techniques include regression,
classification, clustering, decision tree, neural networks, and anomaly
detection.

Regression. (The first machine learning technique uses input data to predict
numerical value)
Classification
Clustering
Decision Tree
Neural Networks
Anomaly Detection
3 Essensitals
Data
• Quantity: Increasing the data used to train the model can improve its accuracy.
• Quality: Using accurate and high-quality data can improve the model’s overall
performance.
• Preprocessing: Cleaning and preprocessing data to handle missing values and
outliers can improve accuracy.
• Augmentation: Expanding the size of real data using data augmentation
techniques can help with smaller datasets and models that experience
overfitting.
Model
• Selection: Experimenting with different model selection techniques to find the best
model for the data.
• Tuning: Adjusting hyperparameters to optimize model performance.
• Ensemble: Combining multiple models for better performance.
Validation
• Cross validation: Testing the accuracy of the model on multiple and diverse subsets
of data.
Classifier:
A classifier in machine learning is an algorithm that automatically
orders or categorizes data into one or more of a set of “classes.” One
of the most common examples is an email classifier that scans emails
to filter them by class label: Spam or Not Spam

Regressior

Any variable in a regression model that is used to predict a response


variable.
ML Regeressior

Mathematical Methods that allow data to predict a continuous outcome


(y) based on the value one or more predictor variables(X)
Model
Performance
Use of Mallows’ Cp,
Akaike’s AIC, and
Amemiya’s PC Akaike’s AIC
•Purpose: AIC is used to compare
different models by quantifying the
trade-off between the goodness of fit
and the complexity of the model.
•When to Use: It is widely used in various
statistical models, including regression,
Mallows’ Cp: Best for selecting time series, and machine learning models.
the subset of predictors in
linear regression models. •Interpretation:
Akaike’s AIC: Versatile and • Lower AIC values indicate a better
widely applicable to various model.
model types, focusing on • AIC penalizes models with more
balancing fit and complexity. parameters to avoid overfitting.
Amemiya’s PC: Specifically
designed for enhancing •Advantage: It is applicable to a wide
predictive accuracy, making it range of models and provides a way to
ideal for predictive modeling. compare models with different numbers
of parameters.
Mallows’ Cp Amemiya’s PC (Prediction Criterion)
•Purpose: Mallows’ Cp is used to assess •Purpose: Amemiya’s PC is used for
the fit of a regression model while model selection with a focus on
accounting for the number of predictors. predictive accuracy.
•When to Use: It is particularly useful •When to Use: It is particularly useful
in linear regression for selecting the when the primary goal is to predict
subset of predictors that provide the future observations accurately.
best fit.
•Interpretation: •Interpretation:
• If Cp is approximately equal to • Lower PC values indicate a
the number of predictors better predictive model.
(including the intercept), the • Like AIC, it balances model fit
model is considered adequate. and complexity but is
• A Cp value much larger than the specifically geared towards
number of predictors indicates prediction.
overfitting, while a value much
smaller suggests underfitting.
•Advantage: It provides a •Advantage: It is tailored for
straightforward measure to balance the situations where prediction is the
trade-off between model complexity and primary concern, providing a direct
goodness of fit. measure of predictive performance.
Model Mallows Cp AIC Amemiya PC
Bagging
4.350833 78.252619 3.614574
KNN 3.495841 77.303012 3.567284
SVM 5.466688 79.470128 3.676292
XGB 16.27504 90.153572 4.274105
Mallows' Cp
•KNN has the lowest Mallows' Cp value (3.495841), suggesting it has the best balance of fit
and number of predictors.
•Bagging follows with a value of 4.350833.
•SVM is next with a value of 5.466688.
•XGB has the highest value (16.27504), indicating potential overfitting or unnecessary
complexity.
AIC (Akaike Information Criterion)
•KNN also has the lowest AIC value (77.303012), indicating it provides the best fit with the
fewest parameters.
•Bagging is close behind with an AIC of 78.252619.
•SVM comes next with an AIC of 79.470128.
•XGB has the highest AIC (90.153572), suggesting it is the least preferred model according
to AIC.
Amemiya's PC
•KNN again has the lowest value (3.567284), indicating the best predictive accuracy.
•Bagging is next with a value of 3.614574.
•SVM follows with a value of 3.676292.
•XGB has the highest value (4.274105), suggesting it has the poorest predictive performance.

You might also like