4/8/25, 7:44 PM 1stTask.

ipynb - Colab

Group Members

1. 21/04905 James Wainaina Githirwa

2. 21/04883 Fabian Ndung'u
3. 21/06700 Peter Kamau
4. 21/04956 Oliver Samwel
5. 21/05462 Purity Njenga
6. 21/05041 Caleb Sirma
7. 21/05119 Bramwel wanyoike
8. ⁠19/02645 Ian Karanja

keyboard_arrow_down Import Libraries

This step imports the necessary libraries for data manipulation, modeling, and visualization.

import pandas as pd
from sklearn.model_selection import train_test_split
from sklearn.ensemble import RandomForestClassifier, GradientBoostingClassifier
from sklearn.tree import DecisionTreeClassifier
from sklearn.metrics import (accuracy_score, classification_report,
confusion_matrix, precision_recall_curve,
PrecisionRecallDisplay)
from imblearn.over_sampling import SMOTE
from sklearn.utils import resample
import seaborn as sns
import matplotlib.pyplot as plt
import warnings

# Suppress FutureWarnings
warnings.filterwarnings("ignore", category=FutureWarning)

keyboard_arrow_down Load and Preprocess Data

This function loads the dataset and performs initial preprocessing, including removing non-predictive columns and handling missing values.

# Load and preprocess data

def load_and_preprocess(filepath):
data = pd.read_csv(filepath)

# Remove non-predictive columns

non_predictive = [
"rapid_flu_results", "rapid_strep_results",
"cxr_findings", "cxr_impression", "cxr_label", "cxr_link",
"batch_date", "test_name", "swab_type"
]
data.drop(columns=[col for col in non_predictive if col in data.columns], inplace=True)

# Handle missing values

data.dropna(subset=["covid19_test_results"], inplace=True)

# Numerical imputation
num_cols = data.select_dtypes(include=['number']).columns
data[num_cols] = data[num_cols].fillna(data[num_cols].mean())

# Categorical imputation
cat_cols = data.select_dtypes(exclude=['number']).columns.drop('covid19_test_results')
for col in cat_cols:
data[col] = data[col].fillna(data[col].mode()[0])

return data

keyboard_arrow_down Handle Class Imbalance

This function balances the dataset by down-sampling the majority class (negative cases) to a specified size.

# Handle class imbalance

def balance_dataset(data):
majority = data[data['covid19_test_results'] == 'Negative']
minority = data[data['covid19_test_results'] == 'Positive']

https://colab.research.google.com/drive/1W1rtHxuRmCvggLRwkGG5j0Ji0cJLpOKf#scrollTo=wjI5rECNSP2q&printMode=true 1/6
4/8/25, 7:44 PM 1stTask.ipynb - Colab
# Downsample majority
majority_down = resample(majority,
replace=False,
n_samples=5000,
random_state=42)

return pd.concat([majority_down, minority], ignore_index=True)

keyboard_arrow_down Main Execution

This section executes the main workflow, including loading, balancing the dataset, and preparing for modeling.

# Main execution
data = load_and_preprocess('coronavirusdataset.csv')
balanced_data = balance_dataset(data)

# Preprocess for modeling

X = pd.get_dummies(balanced_data.drop('covid19_test_results', axis=1), drop_first=True)
y = balanced_data['covid19_test_results'].map({'Negative': 0, 'Positive': 1})

keyboard_arrow_down Split Data

This step splits the data into training and testing sets, ensuring stratification based on the target variable.

# Split data before resampling

X_train, X_test, y_train, y_test = train_test_split(
X, y,
test_size=0.3,
stratify=y,
random_state=42
)

keyboard_arrow_down Apply SMOTE

This step applies SMOTE to the training data to balance the classes further.

# Apply SMOTE only to training data

smote = SMOTE(sampling_strategy='auto', random_state=42)
X_res, y_res = smote.fit_resample(X_train, y_train)

print("\nClass distribution after resampling:")

print(pd.Series(y_res).value_counts())

Class distribution after resampling:

covid19_test_results
0 3500
1 3500
Name: count, dtype: int64

keyboard_arrow_down Model Training and Evaluation

This section defines the models, trains them, and evaluates their performance using accuracy, classification reports, confusion matrices, and
precision-recall curves.

# Model training and evaluation

models = {
"Random Forest": RandomForestClassifier(class_weight='balanced', random_state=42),
"Gradient Boosting": GradientBoostingClassifier(random_state=42),
"Decision Tree": DecisionTreeClassifier(class_weight='balanced', random_state=42)
}

for name, model in models.items():

# Training
model.fit(X_res, y_res)

# Prediction
y_pred = model.predict(X_test)
y_proba = model.predict_proba(X_test)[:,1] if hasattr(model, "predict_proba") else [0]*len(y_test)

# Evaluation
print(f"\n{name} Evaluation:")

https://colab.research.google.com/drive/1W1rtHxuRmCvggLRwkGG5j0Ji0cJLpOKf#scrollTo=wjI5rECNSP2q&printMode=true 2/6
4/8/25, 7:44 PM 1stTask.ipynb - Colab
print("Accuracy:", accuracy_score(y_test, y_pred))
print(classification_report(y_test, y_pred, target_names=['Negative', 'Positive']))

# Confusion Matrix
cm = confusion_matrix(y_test, y_pred)
plt.figure(figsize=(6,4))
sns.heatmap(cm, annot=True, fmt='d', cmap='Blues',
xticklabels=['Predicted Negative', 'Predicted Positive'],
yticklabels=['Actual Negative', 'Actual Positive'])
plt.title(f"{name} Confusion Matrix")
plt.show()

# Precision-Recall Curve
precision, recall, _ = precision_recall_curve(y_test, y_proba)
disp = PrecisionRecallDisplay(precision=precision, recall=recall)
disp.plot()
plt.title(f"{name} Precision-Recall Curve")
plt.show()

# Feature Importance (if available)

if hasattr(model, 'feature_importances_'):
importances = pd.Series(model.feature_importances_, index=X.columns)
top_features = importances.sort_values(ascending=False).head(10)

plt.figure(figsize=(10,6))
top_features.sort_values().plot.barh(color='darkgreen')
plt.title(f"{name} - Top 10 Features")
plt.xlabel("Importance Score")
plt.tight_layout()
plt.show()

https://colab.research.google.com/drive/1W1rtHxuRmCvggLRwkGG5j0Ji0cJLpOKf#scrollTo=wjI5rECNSP2q&printMode=true 3/6
4/8/25, 7:44 PM 1stTask.ipynb - Colab

Random Forest Evaluation:

Accuracy: 0.9714854111405835
precision recall f1-score support

Negative 1.00 0.98 0.99 1500

Positive 0.05 0.25 0.09 8

accuracy 0.97 1508

macro avg 0.52 0.61 0.54 1508
weighted avg 0.99 0.97 0.98 1508

https://colab.research.google.com/drive/1W1rtHxuRmCvggLRwkGG5j0Ji0cJLpOKf#scrollTo=wjI5rECNSP2q&printMode=true 4/6
4/8/25, 7:44 PM 1stTask.ipynb - Colab

Gradient Boosting Evaluation:

Accuracy: 0.9602122015915119
precision recall f1-score support

Negative 1.00 0.96 0.98 1500

Positive 0.07 0.50 0.12 8

accuracy 0.96 1508

macro avg 0.53 0.73 0.55 1508
weighted avg 0.99 0.96 0.98 1508

https://colab.research.google.com/drive/1W1rtHxuRmCvggLRwkGG5j0Ji0cJLpOKf#scrollTo=wjI5rECNSP2q&printMode=true 5/6
4/8/25, 7:44 PM 1stTask.ipynb - Colab

Decision Tree Evaluation:

Accuracy: 0.9602122015915119
precision recall f1-score support

Negative 1.00 0.96 0.98 1500

Positive 0.02 0.12 0.03 8

accuracy 0.96 1508

macro avg 0.51 0.54 0.51 1508
weighted avg 0.99 0.96 0.97 1508

https://colab.research.google.com/drive/1W1rtHxuRmCvggLRwkGG5j0Ji0cJLpOKf#scrollTo=wjI5rECNSP2q&printMode=true 6/6