import gradio as gr
import numpy as np
import pandas as pd
import subprocess
import sys 
subprocess.check_call([sys.executable,'-m','pip','install','tensorflow'])
subprocess.check_call([sys.executable,'-m','pip','install','scikit-learn'])
from sklearn.preprocessing import LabelEncoder, LabelBinarizer
from keras.models import Sequential
from keras.layers import Dense
from keras.wrappers.scikit_learn import KerasClassifier
from sklearn.model_selection import KFold
from sklearn.ensemble import RandomForestClassifier

# Load data
df = pd.read_csv('ExperimentalMigraneData.csv')
X = df[['Age','Duration','Frequency','Location','Character','Intensity','Nausea','Vomit','Phonophobia','Photophobia','Visual','Sensory','Dysphasia','Dysarthria','Vertigo','Tinnitus','Hypoacusis','Diplopia','Visual_defect','Ataxia','Conscience','Paresthesia','DPF', 'On Periods']].values  #selección de variables de entrada
Y = df['Types']  #select target

# Define the base Keras model
def baseline_model():
    model = Sequential()
    model.add(Dense(14, input_dim = 24, activation = 'relu')) # Rectified Linear Unit Activation Function
    model.add(Dense(14, activation = 'relu'))
    model.add(Dense(3, activation = 'softmax')) # Softmax for multi-class classification
    model.compile(loss = 'categorical_crossentropy', optimizer = 'adam', metrics = ['accuracy'])
    return model

# Define the Keras Classifier to use the base model
estimator = KerasClassifier(build_fn = baseline_model, epochs = 100, batch_size = 10, verbose = 0)

# Train the model
estimator.fit(X, Y)

# Define the input component with 24 number inputs
inputs = []
for i in range(24):
    inputs.append(gr.inputs.Number(label=df.columns[i]))

# Define the output component to show the predicted output
output = gr.outputs.Label(label="Output")

# Define the migraine type mapping dictionary
migraine_types = {0: 'Non-Menstrual Migraine',
                  1: 'Menstrual Migraine',
                  2: 'Others'}

# Define the Gradio interface function
def predict(*args):
    # Convert the inputs into a numpy array
    input_array = np.array(args).reshape(1, -1)
    # Use the pre-trained estimator to predict the output based on the input array
    y_pred = estimator.predict(input_array)
    # Map the integer prediction to corresponding migraine type
    predicted_type = migraine_types[int(y_pred[0])]
    # Return the predicted output as text
    return predicted_type

# Run the Gradio interface
interface = gr.Interface(fn=predict, inputs=inputs, outputs=output)
interface.launch()