PFE - Nexvia

Reverse engineering of a determinist model

The main objective of this project is to recover the modeling of the calculations of acquisition costs when buying a property. From a base of training built on a deterministic model our goal will be to find the modeling from machine learning algorithms (reverse engineering).

You can find the slideshow for this project here

Some parts of this project have been purposely deleted for privacy reasons.

Setup and details

• Main scheme : https://goo.gl/hd1Hdr

Install

• Clone the project with git clone url.
• Run . setup.sh.

Get started

• Run the virtualenv with . venv/bin/activate

Within python or ipython shell you can execute commands below to test a basic worflow :

• Basic workflow :

from dev.processing.data_processing import *
from dev.processing.extract import *
from dev.processing.load_data import *
from dev.analysis.analysis import *
from dev.analysis.error import *
from dev.model.model import *
import pandas
import seaborn
import matplotlib.pyplot as pyplot

# Loading raw data
data = load_data()

# Cleaning data (remove na if needed, remove undesired output, etc...)
cleaned_data = cleaning_data(data)

# Processing data (encode qualitative data, select features, etc...)
processed_data = processing_full(cleaned_data)

# Split data set intro train and test sets
x_train, y_train , x_test, y_test = split_dataset(processed_data, 0.3, "output-cumulatedCostsBuy_homeAcquisitionCosts_1")

# Linear Regression (easily modifiable, for exampe with "model_random_forest" for a Random Forest)
model = model_regression(x_train, y_train)

# Predictions
y_pred = model.predict(x_test)

# Print score performances
error_print_score(x_test, y_test, y_pred)

Use run.py interface

• Launch python run.py to get into the interface.

• run.py functional scheme : https://goo.gl/9KgdKF

An end of studies' project made by Aymeric Duchein, Benoit Pimpaud, Jordan Rostren and Paul Schaeffer.