You are previewing Programming Collective Intelligence.

Programming Collective Intelligence

Cover of Programming Collective Intelligence by Toby Segaran Published by O'Reilly Media, Inc.
  1. Programming Collective Intelligence
    1. SPECIAL OFFER: Upgrade this ebook with O’Reilly
    2. A Note Regarding Supplemental Files
    3. Praise for Programming Collective Intelligence
    4. Preface
      1. Prerequisites
      2. Style of Examples
      3. Why Python?
      4. Open APIs
      5. Overview of the Chapters
      6. Conventions
      7. Using Code Examples
      8. How to Contact Us
      9. Safari® Books Online
      10. Acknowledgments
    5. 1. Introduction to Collective Intelligence
      1. What Is Collective Intelligence?
      2. What Is Machine Learning?
      3. Limits of Machine Learning
      4. Real-Life Examples
      5. Other Uses for Learning Algorithms
    6. 2. Making Recommendations
      1. Collaborative Filtering
      2. Collecting Preferences
      3. Finding Similar Users
      4. Recommending Items
      5. Matching Products
      6. Building a del.icio.us Link Recommender
      7. Item-Based Filtering
      8. Using the MovieLens Dataset
      9. User-Based or Item-Based Filtering?
      10. Exercises
    7. 3. Discovering Groups
      1. Supervised versus Unsupervised Learning
      2. Word Vectors
      3. Hierarchical Clustering
      4. Drawing the Dendrogram
      5. Column Clustering
      6. K-Means Clustering
      7. Clusters of Preferences
      8. Viewing Data in Two Dimensions
      9. Other Things to Cluster
      10. Exercises
    8. 4. Searching and Ranking
      1. What's in a Search Engine?
      2. A Simple Crawler
      3. Building the Index
      4. Querying
      5. Content-Based Ranking
      6. Using Inbound Links
      7. Learning from Clicks
      8. Exercises
    9. 5. Optimization
      1. Group Travel
      2. Representing Solutions
      3. The Cost Function
      4. Random Searching
      5. Hill Climbing
      6. Simulated Annealing
      7. Genetic Algorithms
      8. Real Flight Searches
      9. Optimizing for Preferences
      10. Network Visualization
      11. Other Possibilities
      12. Exercises
    10. 6. Document Filtering
      1. Filtering Spam
      2. Documents and Words
      3. Training the Classifier
      4. Calculating Probabilities
      5. A Naïve Classifier
      6. The Fisher Method
      7. Persisting the Trained Classifiers
      8. Filtering Blog Feeds
      9. Improving Feature Detection
      10. Using Akismet
      11. Alternative Methods
      12. Exercises
    11. 7. Modeling with Decision Trees
      1. Predicting Signups
      2. Introducing Decision Trees
      3. Training the Tree
      4. Choosing the Best Split
      5. Recursive Tree Building
      6. Displaying the Tree
      7. Classifying New Observations
      8. Pruning the Tree
      9. Dealing with Missing Data
      10. Dealing with Numerical Outcomes
      11. Modeling Home Prices
      12. Modeling "Hotness"
      13. When to Use Decision Trees
      14. Exercises
    12. 8. Building Price Models
      1. Building a Sample Dataset
      2. k-Nearest Neighbors
      3. Weighted Neighbors
      4. Cross-Validation
      5. Heterogeneous Variables
      6. Optimizing the Scale
      7. Uneven Distributions
      8. Using Real Data—the eBay API
      9. When to Use k-Nearest Neighbors
      10. Exercises
    13. 9. Advanced Classification: Kernel Methods and SVMs
      1. Matchmaker Dataset
      2. Difficulties with the Data
      3. Basic Linear Classification
      4. Categorical Features
      5. Scaling the Data
      6. Understanding Kernel Methods
      7. Support-Vector Machines
      8. Using LIBSVM
      9. Matching on Facebook
      10. Exercises
    14. 10. Finding Independent Features
      1. A Corpus of News
      2. Previous Approaches
      3. Non-Negative Matrix Factorization
      4. Displaying the Results
      5. Using Stock Market Data
      6. Exercises
    15. 11. EVOLVING INTELLIGENCE
      1. What Is Genetic Programming?
      2. Programs As Trees
      3. Creating the Initial Population
      4. Testing a Solution
      5. Mutating Programs
      6. Crossover
      7. Building the Environment
      8. A Simple Game
      9. Further Possibilities
      10. Exercises
    16. 12. Algorithm Summary
      1. Bayesian Classifier
      2. Decision Tree Classifier
      3. Neural Networks
      4. Support-Vector Machines
      5. k-Nearest Neighbors
      6. Clustering
      7. Multidimensional Scaling
      8. Non-Negative Matrix Factorization
      9. Optimization
    17. A. Third-Party Libraries
      1. Universal Feed Parser
      2. Python Imaging Library
      3. Beautiful Soup
      4. pysqlite
      5. NumPy
      6. matplotlib
      7. pydelicious
    18. B. Mathematical Formulas
      1. Euclidean Distance
      2. Pearson Correlation Coefficient
      3. Weighted Mean
      4. Tanimoto Coefficient
      5. Conditional Probability
      6. Gini Impurity
      7. Entropy
      8. Variance
      9. Gaussian Function
      10. Dot-Products
    19. Index
    20. About the Author
    21. Colophon
    22. SPECIAL OFFER: Upgrade this ebook with O’Reilly
O'Reilly logo

The Fisher Method

The Fisher method, named for R. A. Fisher, is an alternative method that's been shown to give very accurate results, particularly for spam filtering. This is the method used by SpamBayes, an Outlook plug-in written in Python. Unlike the naïve Bayesian filter, which uses the feature probabilities to create a whole document probability, the Fisher method calculates the probability of a category for each feature in the document, then combines the probabilities and tests to see if the set of probabilities is more or less likely than a random set. This method also returns a probability for each category that can be compared to the others. Although this is a more complex method, it is worth learning because it allows much greater flexibility when choosing cutoffs for categorization.

Category Probabilities for Features

With the naïve Bayesian filter discussed earlier, you combined all of the Pr(feature | category) results to get an overall document probability, and then flipped it around at the end. In this section, you'll begin by calculating how likely it is that a document fits into a category given that a particular feature is in that document—that is, Pr(category | feature). If the word "casino" appears in 500 documents, and 499 of those are in the bad category, "casino" will get a score very close to 1 for bad.

The normal way to calculate Pr(category | feature) would be:

(number of documents in this category with the feature) / (total number of documents with the feature) ...

The best content for your career. Discover unlimited learning on demand for around $1/day.