import urllib.request
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns

from sklearn.model_selection import train_test_split, RandomizedSearchCV, cross_val_score
from sklearn.preprocessing import StandardScaler, MinMaxScaler
from sklearn.linear_model import LogisticRegression
from sklearn.svm import SVC
from sklearn.tree import DecisionTreeClassifier
from sklearn.ensemble import RandomForestClassifier
from sklearn.neighbors import KNeighborsClassifier
from sklearn.discriminant_analysis import LinearDiscriminantAnalysis, QuadraticDiscriminantAnalysis
from sklearn.naive_bayes import GaussianNB
from sklearn.neural_network import MLPClassifier
from sklearn.model_selection import GridSearchCV
from sklearn.model_selection import cross_val_score, cross_val_predict
from sklearn.metrics import accuracy_score, recall_score, precision_score, f1_score
import joblib
import os
from sklearn.metrics import confusion_matrix
import seaborn as sns
from sklearn.metrics import roc_curve, roc_auc_score

from mlxtend.evaluate import bias_variance_decomp
import gradio as gr

my_model_dict = {}
model_path = '.'

for model_file in os.listdir(model_path):
    tmp_text = os.path.splitext(model_file)
    if len(tmp_text) < 2 or tmp_text[1] != '.joblib':
        continue
    model_filename = os.path.join(model_path, model_file)
    model = joblib.load(model_filename)
    my_model_dict[model_file.split('_')[2]] = model

# 1 example case
examples = [['Logistic Regression', 41, 1, 2, 135.0, 203.0, 0, 1, 132.0, 0, 3.5, 2, 3, 6],
            ['K-Nearest Neighbors', 51, 1, 2, 135.0, 203.0, 0, 1, 132.0, 0, 3.5, 2, 3, 6],
            ['Random Forest', 51, 1, 2, 135.0, 203.0, 0, 1, 132.0, 0, 3.5, 2, 3, 6]]

c_model_name = gr.inputs.Dropdown(choices=['Logistic Regression', 'SVM', 'Decision Tree', 'Random Forest',
                                           'K-Nearest Neighbors', 'Linear Discriminant Analysis',
                                           'Quadratic Discriminant Analysis', 'Naive Bayes', 'Multi-layer Perceptron'],
                                  label='model_name')
c_age = gr.inputs.Slider(minimum=1, maximum=100, step=1, label="age")  # 41
c_sex = gr.inputs.Radio(choices=[0, 1], label="sex")  # 1
c_chest_pain_type = gr.inputs.Dropdown(choices=[1, 2, 3, 4], label="chest_pain_type")  # 2
c_resting_blood_pressure = gr.inputs.Slider(minimum=50, maximum=300, step=0.1, label="resting_blood_pressure")  # 135.0
c_serum_cholesterol = gr.inputs.Slider(minimum=100, maximum=1000, step=0.1, label="serum_cholesterol")  # 203.0
c_fasting_blood_sugar = gr.inputs.Slider(minimum=0, maximum=2, step=1, label="fasting_blood_sugar")  # 0
c_resting_ecg = gr.inputs.Slider(minimum=0, maximum=2, step=1, label="resting_ecg")  # 1
c_max_heart_rate_achieved = gr.inputs.Slider(minimum=50, maximum=300, step=0.01, label="max_heart_rate_achieved")  # 132
c_exercise_induced_angina = gr.inputs.Slider(minimum=0, maximum=1, step=1, label="exercise_induced_angina")  # 0
c_oldpeak = gr.inputs.Slider(minimum=0, maximum=10, step=0.01, label="oldpeak")  # 3.5
c_slope = gr.inputs.Dropdown(choices=[1, 2, 3], label="slope")  # 2
c_num_major_vessels = gr.inputs.Slider(minimum=0, maximum=3, step=1, label="num_major_vessels")  # 3
c_thalassemia = gr.inputs.Dropdown(choices=[3, 6, 7], label="thalassemia")  # 6

outputs = gr.outputs.Textbox()  # returns the top prediction and confidence score


# Define the function for prediction
def predict(model_name, age, sex, chest_pain_type, resting_blood_pressure, serum_cholesterol, fasting_blood_sugar,
            resting_ecg, max_heart_rate_achieved, exercise_induced_angina, oldpeak, slope, num_major_vessels,
            thalassemia):
    # Create the input features
    pred_data = np.array([[age, sex, chest_pain_type, resting_blood_pressure, serum_cholesterol, fasting_blood_sugar,
                           resting_ecg, max_heart_rate_achieved, exercise_induced_angina, oldpeak, slope,
                           num_major_vessels, thalassemia]])
    print(pred_data)
    # Scale the input features
    scaler = StandardScaler()
    scaler.fit(pred_data)
    pred_data_transform = scaler.transform(pred_data)
    # Make the prediction
    model = my_model_dict[model_name]
    setattr(model, 'probability', True)
    prediction = model.predict_proba(pred_data_transform)
    # Return the predicted class probabilities
    return f"Positive: {prediction[0][0]:.2%}\n Negative:{prediction[0][1]:.2%}"

gr.Interface(fn=predict, inputs= [c_model_name, c_age, c_sex, c_chest_pain_type, c_resting_blood_pressure,c_serum_cholesterol,
                                  c_fasting_blood_sugar, c_resting_ecg,c_max_heart_rate_achieved, c_exercise_induced_angina,
                                  c_oldpeak,c_slope, c_num_major_vessels, c_thalassemia],
             outputs=outputs, examples=examples,
             description='Please click one line at examples table.').launch(debug=True)