42 PROJECT: Algorithm re-constituting Harry Potter's Poudlard’s Sorting Hat using Logistic Regression. The model is a logistic regression trained following the gradient descent algorithm.
python3 -m venv venv
venv/bin/pip3 install -r requirements.txt
source venv/bin/activate
Run python describe.py datasets/dataset_train.csv
Mandatory argument: Dataset file
--> Displays the count, mean, standard desviation, min, 25% percentil, median, 75% percentil and max value of all the numerical features (except the index) in the dataset pased as an argument.
Run python describe.py datasets/dataset_train.csv
Mandatory argument: Dataset file
Optional arguments:
- -a: Plot the histogram of all the features
--> Plots the histogram of the course (feature) with the most homogeneous score disribution between all four houses: the course with the smallest standard desviation.
Run python scatter_plot.py datasets/dataset_train.csv
Mandatory argument: Dataset file
Optional arguments:
- -a: Plot the histogram of all the features
--> Plots the scatter plot of the two features that are more similar. The Peason's coeficient of all the pairs of fetures is calculated to find out the ones that are correlated.
Run python pair_plot.py datasets/dataset_train.csv
Mandatory argument:
- Dataset file
--> Plots the scater plot of all the combination of features and the histogram of all of them. From this visualization we can deduce from which features we can extract more informatin to train our model.
Run python logreg_train.py datasets/dataset_train.csv
Mandatory argument: Dataset file
Optional arguments:
- -h: Display help information
- -v {1, 2, 3}: verbosity level
- -lr {float value}: learning rate value (default 0.01)
- -it {int value}: number of iterations to train the model (default 100)
- -tr {float value 0.0-1.0}: defines portion of training set when splitting data into training/testing sets
- -cst: plot of the cost function history
- -ev: display evaluation metrics (accuracy, precision, ...)
- -f: asks user to choose which features to use to train the model
--> Generates a logreg_model.42 file which contains the model and its weights.
Run python3 logreg_predict.py logreg_model.42 datasets/dataset_test.csv
Mandatory argument:
- Dataset file
- Model and weights
--> This will generate a houses.csv file in the datasets/ folder containing all the results of the model cointained in logreg_model.42 prediction for the dataset_train.csv.