Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                
Scribd Logo
Upload

Welcome to Scribd!

  • 9 views

    L5 SubjectReview

    Uploaded by

    Shaiba Shoshi
    The document discusses various machine learning algorithms including linear regression, decision trees, logistic regression, and cluster analysis. It also covers topics like data preparation, exploration, partitioning, and overfitting. Time series forecasting is discussed including partitioning time series data into training and validation periods to evaluate models.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd

    L5 SubjectReview

    Uploaded by

    Shaiba Shoshi
    9 views18 pages
    The document discusses various machine learning algorithms including linear regression, decision trees, logistic regression, and cluster analysis. It also covers topics like data preparation, exploration, partitioning, and overfitting. Time series forecasting is discussed including partitioning time series data into training and validation periods to evaluate models.

    Copyright

    © © All Rights Reserved

    Available Formats

    PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    The document discusses various machine learning algorithms including linear regression, decision trees, logistic regression, and cluster analysis. It also covers topics like data preparation, exploration, partitioning, and overfitting. Time series forecasting is discussed including partitioning time series data into training and validation periods to evaluate models.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    9 views18 pages

    L5 SubjectReview

    Uploaded by

    Shaiba Shoshi
    The document discusses various machine learning algorithms including linear regression, decision trees, logistic regression, and cluster analysis. It also covers topics like data preparation, exploration, partitioning, and overfitting. Time series forecasting is discussed including partitioning time series data into training and validation periods to evaluate models.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    You are on page 1of 18

    Subject Review

    • Data preparation and exploration

    • Data analysis using charts

    • Introduction to Decision Trees

    • Introduction to Cluster Analysis

    • Introduction to Linear Regression

    • Introduction to Logistic Regression

    • Introduction to Forecasting
    Linear Regression

    • supervised machine learning algorithm used for


    prediction tasks
    • To model the relationship between a numerical
    response and one or more explanatory variables

    Yˆ = 2 + 1.25 X
    Decision Tree

    Branch / Sub-
    Root Node
    Tree

    Decision Node Splitting Decision Node

    Terminal Node
    Decision Node Terminal Node Terminal Node
    (Leaf)

    Terminal Node Terminal Node

    • supervised machine learning algorithm used for both classification and prediction tasks
    • graphical representation of a decision-making process
    Logistic Regression

    • Supervised machine learning algorithm used for classifying binary variables


    • The logistic regression model predicts the probability that a given observation belongs to a
    particular class. The model outputs probabilities between 0 and 1
    Cluster Analysis

    X2

    X1
    • unsupervised learning task
    • segment the data into a set of homogeneous clusters of records for the purpose of
    generating insights, discovering patterns, data reduction, etc
    Timeseries Forecasting

    predicting future trends and values in time-series data


    Typical Steps in Data Mining

    1. Define/understand the business requirements


    2. Obtain data
    3. Explore, clean, pre-process data
    4. Reduce the data dimension
    5. Determine DM task
    6. Partition the data (for supervised tasks)
    7. Choose the method
    8. Iterative implementation and tuning
    9. Assess results
    10. Deploy best model
    Data Partitioning and Overfitting

    How well will our prediction or classification model perform when we apply it to new data?

    • The chosen model should be able to generalise beyond the dataset that we have at hand.
    • To assure generalisation, we use the concept of data partitioning and try to avoid overfitting.
    Overfitting

    Overfitting in data mining (and in machine learning more broadly) occurs when a model
    learns the training data too well, capturing not just the underlying patterns but also the
    noise and random fluctuations present in the training set.

    As a result, an overfitted model may perform exceptionally well on the training data but fails
    to generalize to new, unseen data.
    Data Partitioning

    Training Data
    • Typically, the largest partition
    • contains the data used to build the models we are
    examining
    Data Partitioning

    Validation Data
    • used to assess the predictive performance of each
    model so that you can compare models and choose
    the best one
    • In some algorithms, the validation partition may be
    used to tune and improve the model
    Data Partitioning

    Test Data
    • sometimes called evaluation partition is used to assess
    the performance of the chosen model with new data
    Data Partitioning

    Common partition percentage:


    • Train: 70% - 80%
    • Validation: 10% - 15%
    • Test: 10% - 15%

    à balance between having enough data to train a robust


    model and having sufficient validation and test data to
    reliably evaluate its performance
    Data Partitioning

    Selling Price Square Age Condition


    Footage Training data
    9500 1926 30 Good
    80%
    11900 2069 40 Excellent
    124800 1720 30 Good
    135000 1396 15 Mint
    Validation data
    142000 1706 32 Excellent
    145000 1847 38 Mint 10%
    169000 1950 27 Mint
    182000 2323 30 Good
    200000 2285 25 Good Testing data
    210000 3752 17 Good 10%
    … … … …
    Timeseries Forecasting

    • in time series, a random partition would create two time series with “holes”
    • standard forecasting methods cannot handle time series with missing values
    • Solution is partition into two periods:
    • the earlier period is set as the training data
    • the later period is set as the validation data
    • Methods are trained on the earlier training period, and their predictive
    performance assessed on the later validation period
    Timeseries Forecasting

    You might also like