Assignment -
1. Introduction
Statistics can be used to understand data as well as uncover knowledge. Statistics is an established field that looks at the quantification, collection, analysis, interpretation and drawing conclusions from data. The foundations of data mining include statistics as well as other disciplines including databases, machine learning and visualization.
In this section, we will explore two classification techniques that can be used in supervised learning. These techniques include logistic regression and Bayesian classification.
1.1. Logistic Regression: is a very useful (and probably most used) member of a class of models called generalized linear models. Like linear regression, we try to find a best-fit set of parameters and model the relationship between a set of variables/covariates. Unlike linear regression, logistic regression can directly predict values that are restricted to the (0, 1) intervals, such as probabilities. It's the go-to method for predicting probabilities or rates, and like linear regression, the coefficients of a logistic regression model can be treated as advice. Logistic regression predicts the probability that an instance belongs to a specific category- for instance the probability that a flight will be delayed. For a logistic regression classifier we will take the covariates and multiply each one by a weight (coefficient) and then add them up. This result will be put into a logistic function, and we'll get a number between 0 and 1. There are two models of logistic regression, binary logistic regression and multinomial logistic regression. Binary logistic regression is typically used when the dependent variable is dichotomous and the independent variable is continuous or categorical. When the dependent variable is not dichotomous and is comprised of more than two categories, multinomial logistic regression can be employed.
1.2. Bayesian Classification: Bayesian classifiers can be used to predict the probability of class membership. Despite violation of the Naïve Bayes requirement of independence among explanatory variables, these classifiers have been found to perform well with relatively small amount of training data. Naïve Bayes classifiers can use continuous and categorical independent variables.
Examples of the use of Bayesian classification, among many others, include those in spam detection, sentiment analysis, and the probability if a visitor to the site would place an order. Bayesian classification allows incorporating new information to update prior probabilities.
There are a number of different implementation of the Bayes rule - for example some algorithms are more appropriate when frequency of words in a text document is important (e.g. when conducting topic categorization) and others when mere occurrence of certain terms is sufficient (e.g., in sentiment analysis).
2. Steps to Completion
For each study the general procedure is to:
- Review theoretical background based on available resources in the course content
- Choose one of the two statistical data mining techniques - logistic regression or Bayesian classification
- Select a dataset from the module's recommended datasets list
- Run an analysis, perform evaluation, and capture the results
- Document your findings and analysis in a data mining analytical report
3. Deliverables
Submit your analysis report by addressing the following critical areas:
Introduction: give some background and context about the domain of application, provide the rationale for the type of analysis, and state the objective clearly.
Analysis: describe the data both qualitatively and quantitatively through exploratory analysis, perform necessary preprocessing activities, give some intuition about the algorithm and core parameters, demonstrate the model building steps along with parameter tuning, and explain all your assumptions.
Result: explain the result and interpret the model output using terms that reflect the application area, perform model evaluation using the appropriate metrics, and leverage visualization.
Conclusion: summarize your main findings, discuss experimental limitations related to the data and/or implementation of the algorithm, and suggest improvement areas as a potentiation future work.
Miscellaneous:
- Proof read your report for correct structure, grammar, and spelling
- Follow appropriate APA formatting and provide all references
- Include your R script and extended model outputs in an Appendix section.
The length of the report should be 7-10 pages excluding the title page, appendix and R script.