Analyze+Model-P1.md

Analyze + Mode Part 1

Analyze & Model - the process of understanding, diagnosing, and refining a machine learning model with the help of interactive visualization

Regression Models

Linear Regression Model

Linear Regression - a supervised machine learning algorithm where the predicted output is continuous and has a constant slope

Logistic Regression Model

Logistic Regression - a classification algorithm used to assign observations to a discrete set of classes. Unlike linear regression which outputs continuous number values, logistic regression transforms its output using the logistic sigmoid function to return a probability value which can then be mapped to two or more discrete classes.

Examples of Categorical Data

• Binary (Pass/Fail)
• Multi (Cats, Dogs, Sheep)
• Ordinal (Low, Medium, High)

Sigmoidal (function) curve

Testing model performance

• Train-test split
• Cross Validation (good for Hyperparameterized tuning)
  • K Fold Cross Validation - ???

Cross Validation

1. Split the data into 3 parts (Training Set => Train, Crossvalidation set = Validate Training data Set, Test Set)

Model tuning

Pick a model that you are getting the best performance and attempt to tune it with regularization and Hyperparameters

• Underfitting vs overfitting
• Regularization
• Hyperparameter tuning
• Cross validation

Feature Engineering

• Feature normalization

Model Tuning - the process of maximizing a model's performance without overfitting or creating too high of a variance

Underfitting - is unable to capture the relationship between the input and output variables accurately

Overfitting - when a statistical model fits exactly against its training data. You can avoid this with regularization. - Regularization - a technique used for tuning the function by adding an additional penalty term in the error function Regularization - - Hyperparameter - is a parameter whose value is used to control the learning process. (Example C, penalty, ) - Penalty - C

Hyperparameter Optimization -

Use Crossvalidation to test you regularization hyperparameter tuning

Feature Normalization - Feature Standardization -

https://ml-cheatsheet.readthedocs.io/en/latest/logistic_regression.html

---

Analyze+Model-P2.md

Analyze + Mode Part 2

Predictive Modeling with Machine Learning

Terminology

---

Training data - data you use to train a machine learning alogrithm - Input features data (Predictors) - Output features data (Response, Outcome)

Supervised data - training data that contains Input and Output features

Supervised Learning Problems - is a machine learning task of learning a function that maps an input to an output data based on example input-output pairs (ie Supervised data).

Examples of Supervised Learning models

• Regression model - Output is a continuous variable | requires the prediction of a quantity
• Classification model - requires that examples be classified into one of two or more classes. Example Titanic dataset Survived (1) or Not Survived (0)

Unsupervised data - training data that only contains Input features data

Unsupervisored Learning Problems - the goal is to identify meaningful patterns in the training data which doesn't contain output data

Examples of Unsupervised Learning models

• K-means clustering
• Reinforcement learning

Classification Task - requires that examples be classified into one of two or more classes. Example Titanic dataset Survived (1) or Not Survived (0)

Types of Classification Tasks

• Binary Classification Problem - Judge by accuracy, other precision, recall
• Multi-label Classification
• Classification Predicitive

Determining Performance

Confusion Matrix - is a table that is often used to describe the performance of a classification model. Use the following table to determine metric performanance.

                     Confusion Matrix

+-----------------+---------------------+---------------------+
|                 | Predicted Negative  | Predicted Positive  |
+-----------------+---------------------+---------------------+
| Actual Negative | True Negative (TN)  | False Positive (FP) |
+-----------------+---------------------+---------------------+
| Actual Postive  | False Negative (FN) | True Positive (TP)  |
+-----------------+---------------------+---------------------+

Examples of Performance metrics

• Precision (aka Positive Predictive Value) - What proportion of positive identifications was actually correct?

  Precision = TP/ (TP + FP)
  

• Recall - What proportion of actual positives was identified correctly?

  Recall = TP/ (TP + FN)
  

• Accuracy - the percentage of correct predictions for the test data

  Accuracy = (TP + TN) / (TP + TN + FP + FN)
  

• Negative Predictive Value (NPV) - ?

  NPV = TN/(TN + FN)
  

• Sensitivity - True Positive Rate. The proportion of actual negatives, which got predicted correctly

  Sensitivity = TP/(TP+TN)
  

• Specifictiy - True Negative Rate. The proportion of actual negatives, which got predicted correctly

  Specifity = TN/(TN + FP)
  

Evaluating Your Classifier

If you are given supervised test data you can perform an 80% 20% split in order to train your alogrithm.

1. Split supervised data into 80/20 split
2. Use the 80% supervised data to train a model
3. Use the 20% supervised data to test your tranined model
4. Compare your Predicted Output of you 20% supervised data input to the actual values and determine your score (Precision, Recall)

---

This idea of a feature is a detail of the data

Statistical modeling is a subfield of mathematics that seeks out relationships between variables in order to predict an outcome.

Classifier - a type of machine learning algorithm used to assign a class label to a data input

• Logistic Regression
• Support Vector machine
• Neural Networks
• Random forest

Analyze-Model.png

Classifier-Evaluation.png

Cross-Validation.png

Data-Extraction.md

Data-Organization.md

Data-Science-Cycle-Overview.png

Data-Science-Overview.md

Linear-Regression.png

Logistic-Regression.png

ML-Templates.md

Organize-Step.png

Sigmoid.png