Page MenuHomec4science
Files F96088627

main.py
No OneTemporary
Actions

File Metadata

Created
Sun, Dec 22, 12:16

main.py
View Options

#main.py
from data_processing_AMP import *
from data_processing_VIAB import *
import pickle
import pandas as pd
from sklearn.model_selection import train_test_split
from sklearn.metrics import classification_report, accuracy_score, roc_auc_score
from xgboost import XGBClassifier
import optuna
from optuna.samplers import TPESampler
import matplotlib.pyplot as plt
from imblearn.over_sampling import SMOTE
import numpy as np
from sklearn.preprocessing import StandardScaler
#Define the used features manually
selected_features = [
'min_trapping', 'max_trapping', 'mean_trapping', 'std_trapping', 'q25_trapping'
]
def augment_data(X, y):
"""
Augments the dataset using SMOTE to balance classes and adds random noise for further augmentation.
Args:
X (pd.DataFrame): Feature set.
y (pd.Series): Target variable.
Returns:
tuple: Augmented feature set and target variable.
"""
smote = SMOTE(random_state=42)
X_res, y_res = smote.fit_resample(X, y)
noise_factor = 0.05
X_res += noise_factor * np.random.normal(loc=0.0, scale=1.0, size=X_res.shape)
return X_res, y_res
def objective(trial):
"""
Objective function for optimizing XGBoost hyperparameters using Optuna.
Args:
trial (optuna.trial.Trial): A trial object that suggests hyperparameter values.
Returns:
float: ROC AUC score of the model on the validation set.
"""
param = {
'n_estimators': trial.suggest_int('n_estimators', 50, 500),
'max_depth': trial.suggest_int('max_depth', 3, 10),
'learning_rate': trial.suggest_float('learning_rate', 0.01, 0.3),
'subsample': trial.suggest_float('subsample', 0.6, 1.0),
'colsample_bytree': trial.suggest_float('colsample_bytree', 0.6, 1.0),
'gamma': trial.suggest_float('gamma', 0, 10),
'reg_alpha': trial.suggest_float('reg_alpha', 0, 1.0),
'reg_lambda': trial.suggest_float('reg_lambda', 1.0, 10.0),
'random_state': 42,
'use_label_encoder': False,
'eval_metric': 'logloss'
}
model = XGBClassifier(**param)
model.fit(X_train_VIAB_aug[selected_features], y_train_VIAB_aug)
y_pred = model.predict(X_test_VIAB_norm[selected_features])
# Return the ROC AUC score for evaluation
return roc_auc_score(y_test_VIAB, model.predict_proba(X_test_VIAB_norm[selected_features])[:, 1])
if __name__ == '__main__':
#in this code we will first train and test a model to predict death on the labeled dataset viability
#then we will use this same model to predict the dataset with unknown death labels : AMP and plot a bar plot to visualize
print("Loading data from pickle files...")
with open('docs_analysed.pkl', 'rb') as f:
docs_analyzed_VIAB = pickle.load(f)
with open('docs_meas.pkl', 'rb') as f:
docs_meas_VIAB = pickle.load(f)
with open('docs_analysed_AMP.pkl', 'rb') as f:
docs_analyzed_AMP = pickle.load(f)
with open('docs_meas_AMP.pkl', 'rb') as f:
docs_meas_AMP = pickle.load(f)
print("Data loaded successfully.")
norm_docs_VIAB = extract_features_VIAB(docs_analyzed_VIAB)
print('first dataset', norm_docs_VIAB.iloc[0])
norm_docs_AMP = normalize_docs_AMP(docs_analyzed_AMP, docs_meas_AMP)
print('second dataset', norm_docs_AMP.iloc[0])
X_VIAB = norm_docs_VIAB.drop(columns=['dead'])
y_VIAB = norm_docs_VIAB['dead']
# Perform a stratified split to maintain the same ratio of classes in both training and testing sets
X_train_VIAB, X_test_VIAB, y_train_VIAB, y_test_VIAB = train_test_split(
X_VIAB, y_VIAB, test_size=0.3, stratify=y_VIAB, random_state=42
)
# Normalize the data
scaler = StandardScaler()
X_train_VIAB_norm = scaler.fit_transform(X_train_VIAB[selected_features])
X_test_VIAB_norm = scaler.transform(X_test_VIAB[selected_features])
X_norm_docs_AMP = scaler.transform(norm_docs_AMP[selected_features])
# Convert back to DataFrame for easier handling
X_train_VIAB_norm = pd.DataFrame(X_train_VIAB_norm, columns=selected_features)
X_test_VIAB_norm = pd.DataFrame(X_test_VIAB_norm, columns=selected_features)
X_norm_docs_AMP = pd.DataFrame(X_norm_docs_AMP, columns=selected_features)
print("Training set class distribution:\n", y_train_VIAB.value_counts())
print("Test set class distribution:\n", y_test_VIAB.value_counts())
# Data Augmentation
X_train_VIAB_aug, y_train_VIAB_aug = augment_data(X_train_VIAB_norm, y_train_VIAB)
# Use Optuna to find the best hyperparameters for XGBoost
#sampler = TPESampler(seed=42) # Use TPE for efficient sampling
study = optuna.create_study(direction='maximize')
study.optimize(objective, n_trials=150, timeout=1000) # Adjust n_trials and timeout as needed
print("Best hyperparameters:", study.best_params)
# Train the best model with the found hyperparameters and augmented data
best_model = XGBClassifier(**study.best_params)
best_model.fit(X_train_VIAB_aug[selected_features], y_train_VIAB_aug)
# Evaluate the best model
y_pred_VIAB = best_model.predict(X_test_VIAB_norm[selected_features])
print("Accuracy:", accuracy_score(y_test_VIAB, y_pred_VIAB))
print("Classification Report:\n", classification_report(y_test_VIAB, y_pred_VIAB))
# Predict the 'dead' status for the second dataset
predicted_dead = best_model.predict(X_norm_docs_AMP[selected_features])
# Add the predicted dead status back to the DataFrame
norm_docs_AMP['predicted_dead'] = predicted_dead
# Combine with antibiotics_quantity
antibiotics_quantity = norm_docs_AMP['antibiotics_quantity']
norm_docs_AMP['antibiotics_quantity'] = antibiotics_quantity
print(norm_docs_AMP)
# Group by antibiotics_quantity and count the number of 0s and 1s
death_counts = norm_docs_AMP.groupby(['antibiotics_quantity', 'predicted_dead']).size().unstack(fill_value=0)
death_counts.plot(kind='bar', stacked=False, color=['blue', 'yellow'])
plt.xlabel('Antibiotics Quantity')
plt.ylabel('Number of Samples')
plt.title('Predicted Death vs Alive by Antibiotics Quantity')
plt.legend(['Alive (0)', 'Dead (1)'])
plt.savefig('bar_plot.png')
plt.close()

Event Timeline

c4science · Help