app.py

# -*- coding: utf-8 -*-
"""logistic_regression.ipynb

Automatically generated by Colab.

Original file is located at
    https://colab.research.google.com/drive/18WbnQFC9t31wB5Ij0LJnr7Xf6Z8X8o2f

**Building a Classifier from Scratch** <br>
**Author: Jigyasu**
"""

import numpy as np

sigmoid = lambda x : 1/(1+np.exp(-1*x))

def cost_function(X, y, w, b):
    m = len(X)
    f = sigmoid(np.dot(X, w) + b)
    # Clipping f to avoid log(0)
    epsilon = 1e-15
    f = np.clip(f, epsilon, 1 - epsilon)
    loss = (-1 * y * np.log(f)) - ((1 - y) * np.log(1 - f))
    cost = np.sum(loss) / m
    return cost

def gradient(X, y, w, b):
    m, n = X.shape
    dw = np.zeros((n,))
    db = 0.
    for i in range(m):
        f = sigmoid(np.dot(X[i], w) + b)
        db += (f - y[i])
        for j in range(n):
            dw[j] += (f - y[i]) * X[i, j]
    dw /= m
    db /= m
    return dw, db

def fit(X, y, alpha=0.01, epochs=1000):
  m, n = X.shape
  w = np.zeros((n, ))
  b = 0.
  for iteration in range(epochs):
    dw, db = gradient(X, y, w, b)
    w -= alpha * dw
    b -= alpha * db
    cost = cost_function(X, y, w, b)
    if iteration % 100 == 0:
      print(f"Epoch: {iteration}, Cost: {cost}")
  return w, b

def predict(X, w, b):
  f = sigmoid(np.dot(X, w) + b)
  predictions = np.array([1 if f[i] > 0.6 else 0 for i in range(len(f))])
  return predictions

def calculate_accuracy(y_true, y_pred):
    correct_predictions = np.sum(y_true == y_pred)
    accuracy = correct_predictions / len(y_true)
    return accuracy

"""**Applying the Model to a Heart Failure Clinical Record Dataset**"""

import pandas as pd

data = pd.read_csv("heart_failure_clinical_records.csv")
X = data.drop(columns=["DEATH_EVENT"]).values
y = data["DEATH_EVENT"].values

#sklearn is used to split and scale the data
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import StandardScaler
scaler = StandardScaler()
X = scaler.fit_transform(X)
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)


w, b = fit(X_train, y_train)

predictions = predict(X_test, w, b)
accuracy = calculate_accuracy(y_test, predictions)
print(f"Accuracy: {accuracy * 100:.2f}%")

import gradio as gr

# Gradio interface
def predict_single_instance(age, anaemia, creatinine_phosphokinase, diabetes, ejection_fraction, high_blood_pressure, platelets, serum_creatinine, serum_sodium, sex, smoking, time):
    # Prepare input
    input_data = np.array([age, anaemia, creatinine_phosphokinase, diabetes, ejection_fraction, high_blood_pressure, platelets, serum_creatinine, serum_sodium, sex, smoking, time]).reshape(1, -1)
    input_data = scaler.transform(input_data)
    # Predict
    prediction = predict(input_data, w, b)
    return "Survived" if prediction[0] == 0 else "Death",

inputs = [
    gr.Number(label="Age"),
    gr.Number(label="Anaemia (0 for positive, 1 for negative)"),
    gr.Number(label="Creatinine Phosphokinase"),
    gr.Number(label="Diabetes (0 for positive, 1 for negative)"),
    gr.Number(label="Ejection Fraction"),
    gr.Number(label="High Blood Pressure (0 for positive, 1 for negative)"),
    gr.Number(label="Platelets"),
    gr.Number(label="Serum Creatinine"),
    gr.Number(label="Serum Sodium"),
    gr.Number(label="Sex (0 for Woman, 1 for Man)"),
    gr.Number(label="Smoking (0 for positive, 1 for negative)"),
    gr.Number(label="Time (Follow-up Period)")
]

outputs = gr.Textbox(label="Prediction")

examples = [
    [75, 0, 582, 0, 20, 1, 265000, 1.9, 130, 1, 0, 4],
    [55, 1, 7861, 1, 38, 0, 263358, 1.1, 137, 1, 0, 6],
    [65, 0, 146, 1, 20, 1, 162000, 2.7, 116, 0, 0, 7]
]

gr.Interface(
    fn=predict_single_instance,
    inputs=inputs,
    outputs=outputs,
    examples=examples,
    title="Heart Stroke Survival Prediction"
).launch()