A small tutorial given at LA DataCon on a basic Data Science. The tutorial was done in python but I wanted to try to do it in R.
What happened and why it happened is what Data Analysis amounts to.
What will happen tomorrow is predictive analytics.
-
Understand the problem and what the expected product is: A bank wants to determine which customers are most likely to need a bank loan. Using data on hand: the final product should be a predictive model that can determine if a new client needs a loan.
-
What data do we have? The data consists of generale features a financial institution would have on its clients. (Ex. Loan amount, Gender, Education,etc) Dataset is mixed with both numeric and categorical data.
-
Analysis and Preparation
- Initial summary showed missing values in several columns of the dataset.
- These missing values were replaced with either the mean for numeric or the mode for categorical data.
- One hot encoding was used on categorical variables.
- A correlation matrix was used to determine which features had the highest correlation.
- Model Selction
- Logistic Regression
- Decision Tree
- Results
- Decision Tree
| Accuracy | Precision | Sensitivity | Specificity |
|---|---|---|---|
| 0.8366013 | 0.9339623 | 0.9339623 | 0.6170213 |
References
- Intro to Data Science, LA DataCon, Zia Khan