Merge pull request #21 from 01-edu/project5

Project5

one_md_per_day_format/projects/project5/project5.md

@@ -0,0 +1,112 @@
+# Credit scoring
+
+The goal of this project is to implement a scoring model based on various source of data (check data documentation) that returns the probability of default. In a nutshell, credit scoring represents an evaluation of how well the bank's customer can pay and is willing to pay off debt. It is also required that you provide an explanation of the score. For example, your model returns that the probability that one client doesn't pay back the loan is very high (90%). The reason behind is that variable_xxx which represents the ability to pay back the past loan is low. The output interpretability will appear in a visualization.
+
+The ability to understand ... . is important. Credit scoring is subject to more and more regulation. Transparency is key. And more generally, more and more companies prefer transparency to black box models.
+
+## Resources
+
+Historical timeline of machine learning techniques applied to credit scoring
+
+- https://hal.archives-ouvertes.fr/hal-02507499v3/document
+- https://www.kaggle.com/c/home-credit-default-risk/data
+
+# Deliverables
+
+## Scoring model
+
+The are 3 expected deliverables associated with the scoring model:
+
+- An exploratory data analysis notebook that describes the insights you find out in the data set.
+- The trained machine learning model with the features engineering pipeline:
+
+  - Do not forget: **Coming up with features is difficult, time-consuming, requires expert knowledge. ‘Applied machine learning’ is basically feature engineering.**
+  - The model is validated if the **AUC on the test set is higher than 75%**.
+  - The labelled test data is not publicly available. However a Kaggle competition uses the same data. The procedure to evaluate test set submission is the same as the one used for the project 1.
+
+### Kaggle submission
+
+The way the Kaggle platform works is explained in the challenge overview page. If you need more details,  I suggest this resource that gives detailed explanations.
+
+- https://towardsdatascience.com/getting-started-with-kaggle-f9138b35ae18
+
+- Create a username following that structure: username_01EDU_ location_MM_YYYY. Submit the description profile and push it on GitHub the first day of the week. Do not touch this file anymore.
+
+- A text document that describes the methodology used to train the machine learning model:
+  - Algorithm
+  - Why the accuracy shouldn't be used in that case ?
+  - Limit and possible improvements
+
+## Model interpretability
+
+This part hasn't been covered during the piscine. Take the time to understand this key concept.
+There are different level of transparency:
+
+- **Global**: understand important variables in a model. This answers the question: "What are the key variables to the model ? ". In that case it will tell if the revenue is more important than the age to the model for example. This allows to check that the model relies on important variables. No one wants his credit to be refused because of the weather in Lisbon !
+- **Local**: each observation gets its own set of interpretability factors. This greatly increases its transparency. We can explain why a case receives its prediction and the contributions of the predictors. Traditional variable importance algorithms only show the results across the entire population but not on each individual case. The local interpretability enables us to pinpoint and contrast the impacts of the factors.
+
+There are 2 tools you can use to analyse your model and its predictions: - Features importance (available if you use a Scikit Learn model) - [SHAP library](https://towardsdatascience.com/explain-your-model-with-the-shap-values-bc36aac4de3d)
+
+Implement a program that takes as input the trained model, the customer id ... and returns:
+
+- the score and the SHAP force plot associated with it
+- Plotly visualisations that show:
+  - key variables describing the client and its loan(s)
+  - comparison between this client and other clients
+
+Choose the 3 clients of your choice, compute the score, run the visualizations on their data and save them.
+
+- Take 2 clients from the train set:
+  - 1 on which the model is correct and the other on which the model is wrong. Try to understand why the model got wrong on this client.
+- Take 1 client from the test set
+
+### Optional
+
+Implement a dashboard (using Dash) that takes as input the customer id and that returns the score and the required visualizations.
+
+- https://stackoverflow.com/questions/54292226/putting-html-output-from-shap-into-the-dash-output-layout-callback
+
+## Deliverables
+
+```
+project
+│   README.md
+│   environment.yml
+│
+└───data
+│   │   ...
+│
+└───results
+│   │
+|   |───model (free format)
+│   │   │   my_own_model.pkl
+│   │   │   model_report.txt
+│   │
+|   |feature_engineering
+│   │   │   EDA.ipynb
+│   │
+|   |───clients_outputs
+|   |   |   client1_correct_train.pdf  (free format)
+│   │   │   client2_wrong_train.pdf  (free format)
+│   │   │   client_test.pdf   (free format)
+│   │
+|   |───dashboard (optional)
+|   |   |   dashboard.py  (free format)
+│   │   │   ...
+|
+|───scripts (free format)
+│   │   train.py
+│   │   predict.py
+│   │   preprocess.py
+```
+
+- `README.md` introduces the project and shows the username.
+- `environment.yml` contains all libraries required to run the code.
+- `username.txt` contains the username, the last modified date of the file **has to correspond to the first day of the project**.
+- `EDA.ipynb` contains the exploratory data analysis. This file is should contain all steps of data analysis that contributed or not to improve the score of the model. It has to be commented so that the reviewer can understand the analysis and run it without any problem.
+
+- `scripts` contains python file(s) that perform(s) the feature engineering, the model's training and prediction on the test set. It could also be one single Jupyter Notebook. It has to be commented to help the reviewers understand the approach and run the code without any bugs.
+
+## Useful resources
+
+- https://towardsdatascience.com/interpretability-in-machine-learning-70c30694a05f

one_md_per_day_format/projects/project5/project5_correction.md

@@ -0,0 +1,45 @@
+# Credit scoring
+
+## Machine learning model
+
+- the model is trained only the training set 
+- the AUC on the test set is higher than 75% 
+- the model learning curves should prove that the model is not overfitting 
+- the training has been stopped early enough to avoid the overfitting
+- the text document should describe the methodology used to train the machine learning model.
+- predict.py runs without any error and returns: 
+
+
+    ```prompt 
+    python predict.py
+
+    AUC on test set: 0.76
+
+    ```
+
+This article gives a complete example of a good modelling approach: 
+
+https://medium.com/thecyphy/home-credit-default-risk-part-2-84b58c1ab9d5
+
+
+## Model's interpretability
+
+- Feature importance:
+    - the importance of all features used by the model are computed and showed in a visualisation. You should be careful here to associate the right variables to the their feature importance. Sometimes, the preprocessing pipeline can remove some features during the features selection step for instance. 
+
+
+- Descriptive variables: 
+These are important to understand for example the age of the client. If the data could be scaled or modified in the preprocessing pipeline but the data visualised here should be "raw". This part is validated if the visualisations are computed for the 3 clients.
+    - visualisations that show at least 10 variables describing the client and its loan(s)
+    - visualisations that show the comparison between this client and other clients.
+
+- SHAP values on the model:
+    - a summary plot shows the important features and their impact on the target. This is optional if you have already computed the features importance. 
+
+- SHAP values on predictions:
+This part is validated if the SHAP visualisations are computed for the 3 clients.
+    - a force plot shows what variables contributes the most to the score. 
+    - **check that the score outputted by the force plot corresponds to the one outputted by the model.** 
+
+
+

one_md_per_day_format/projects/project5/readme_data.md

@@ -0,0 +1,46 @@
+# Credit scoring data description
+
+This file describes the available data for the project.
+
+![alt data description](project5_data_description.png "Credit scoring data description")
+
+## application_{train|test}.csv
+
+This is the main table, broken into two files for Train (with TARGET) and Test (without TARGET).
+Static data for all applications. One row represents one loan in our data sample.
+
+## bureau.csv
+
+All client's previous credits provided by other financial institutions that were reported to Credit Bureau (for clients who have a loan in our sample).
+For every loan in our sample, there are as many rows as number of credits the client had in Credit Bureau before the application date.
+
+## bureau_balance.csv
+
+Monthly balances of previous credits in Credit Bureau.
+This table has one row for each month of history of every previous credit reported to Credit Bureau – i.e the table has (#loans in sample * # of relative previous credits * # of months where we have some history observable for the previous credits) rows.
+
+## POS_CASH_balance.csv
+
+Monthly balance snapshots of previous POS (point of sales) and cash loans that the applicant had with Home Credit.
+This table has one row for each month of history of every previous credit in Home Credit (consumer credit and cash loans) related to loans in our sample – i.e. the table has (#loans in sample * # of relative previous credits * # of months in which we have some history observable for the previous credits) rows.
+
+## credit_card_balance.csv
+
+Monthly balance snapshots of previous credit cards that the applicant has with Home Credit.
+This table has one row for each month of history of every previous credit in Home Credit (consumer credit and cash loans) related to loans in our sample – i.e. the table has (#loans in sample * # of relative previous credit cards * # of months where we have some history observable for the previous credit card) rows.
+
+## previous_application.csv
+
+All previous applications for Home Credit loans of clients who have loans in our sample.
+There is one row for each previous application related to loans in our data sample.
+
+
+## installments_payments.csv
+
+Repayment history for the previously disbursed credits in Home Credit related to the loans in our sample.
+There is a) one row for every payment that was made plus b) one row each for missed payment.
+One row is equivalent to one payment of one installment OR one installment corresponding to one payment of one previous Home Credit credit related to loans in our sample.
+
+## HomeCredit_columns_description.csv
+
+This file contains descriptions for the columns in the various data files.