Machine Learning
Fundamentals
In this session, we will define basic terminologies
used in machine learning including:
• Labels
• Algorithms
• Building Predictive model
• Model evaluation metrics
• Feature Engineering
What is a Label?
1 Definition 2 Importance 3 Examples
In machine learning, a Labels are crucial for Examples of labels
label is the target supervised learning, include whether an email
variable that a model is where the model is is spam or not, whether
trying to predict. It is the trained on labeled data a loan applicant will
known outcome or to learn the relationship default, or the price of a
classification for a given between inputs and home.
set of input data. outputs.
Understanding Algorithms
Definition Types Importance
A machine learning algorithm Common algorithms include The choice of algorithm is
is a step-by-step procedure linear regression, decision critical, as it determines the
that a computer uses to learn trees, random forests, and model's ability to capture
from data and make neural networks, each with patterns and make accurate
predictions or decisions. their own strengths and use predictions from the data.
cases.
Building a Prediction Model
Data Collection Model Evaluation
Gathering relevant and high-quality data is The model's performance is evaluated using
the foundation for building an effective various metrics to ensure it meets the desired
prediction model. level of accuracy.
1 2 3
Model Training
The model is trained on the data, learning to
recognize patterns and make predictions.
Supervised vs. Unsupervised Learning
Supervised Learning Unsupervised Learning Key Differences
In supervised learning, the Unsupervised learning Supervised learning is more
model is trained on labeled involves finding patterns and common and typically leads
data, learning to map inputs structure in unlabeled data to more accurate models,
to known outputs. without explicit guidance. while unsupervised learning is
useful for discovering hidden
insights in data.
Feature Engineering
What is it? Importance
Feature engineering is the process of Feature engineering can significantly impact
creating new input features from raw data to a model's ability to capture relevant patterns
improve a model's performance. and make accurate predictions.
Examples Challenges
Examples include transforming data, creating Feature engineering requires domain
new features, and selecting the most knowledge and can be time-consuming, but
informative features. it's a crucial step in building effective machine
learning models.
Model Evaluation Metrics
Accuracy Precision Recall F1-Score
Measures the Measures the Measures the Combines precision
proportion of correct proportion of true proportion of true and recall into a
predictions made by positives among all positives that the single metric that
the model. positive predictions. model correctly balances both.
identified.
Real-World Applications of Machine
Learning
Healthcare Finance Retail Transportation
Predicting disease Fraud detection, Demand Self-driving cars,
risk, detecting credit risk forecasting, product traffic prediction,
anomalies in assessment, stock recommendation, route optimization,
medical scans, and price prediction, and customer and predictive
personalizing portfolio segmentation, and maintenance.
treatment. optimization. pricing optimization.
