# %%writefile app.py
import joblib
import pandas as pd
import numpy as np
import streamlit as st
from sklearn.preprocessing import LabelEncoder
from lime.lime_tabular import LimeTabularExplainer

# -------------------------------------------------------------------------------------
df = pd.read_csv("combined.csv", index_col=0)
df2 = df.copy()
obj_columns = df.select_dtypes(include=['object']).columns
num_columns = df.select_dtypes(include='number').columns
le_dict = {}
classes_dict = {}
for col in obj_columns:
    le = LabelEncoder()
    df2[col] = le.fit_transform(df[col])
    le_dict[col] = le
    classes_dict[col] = le.classes_
df2['G1'] = df2.pop('G1')
df2['G2'] = df2.pop('G2')
df2['G3'] = df2.pop('G3')
X = df2.iloc[:,:-1]
y = df2.iloc[:,-1]
allCol = X.columns

# -------------------------------------------------------------------------------------
# Load the model from the file
joblib_file = "xgb_model.joblib"
loaded_model = joblib.load(joblib_file)

# -------------------------------------------------------------------------------------
variableExpl = []
with open('student.txt', 'r', encoding='utf-8') as file:
    for line in file:
        cleaned_line = line.strip()
        # Append each cleaned line as a row to the list
        variableExpl.append(cleaned_line)

variableExpl.pop(0)
for i in range(5):
    variableExpl.pop(-1)
for i in range(2):
    variableExpl.pop(-3)

variableExplDict = {}
for i in variableExpl:
    variableExplDict[i.split()[1]] = i

# -------------------------------------------------------------------------------------
def predict_score(inputs):
    if any(value == '' for value in inputs):
        return "Please enter all the inputs."
        
    #-------------------------------------------------------------------------------------------
    # Create a dictionary for each input
    input_df = pd.DataFrame(np.array(inputs).reshape(1, -1), columns=allCol)


    #-------------------------------------------------------------------------------------------
    # label encode each input
    for col in obj_columns:
        if col in input_df.columns:
             input_df[col] = le_dict[col].transform(input_df[col])
    
    #-------------------------------------------------------------------------------------------
    # Make predictions
    pred = loaded_model.predict(input_df)

    # Ensure all columns are numeric
    input_df = input_df.astype(float)
    #-------------------------------------------------------------------------------------------
    # Create a LIME explainer
    explainer = LimeTabularExplainer(training_data=X.values, mode="regression", feature_names=allCol, verbose=True)

    exp = explainer.explain_instance(data_row=input_df.iloc[0].to_numpy(), predict_fn=loaded_model.predict, num_features=33)

    impacts = {}
    for item in exp.as_list():
        impacts[item[0]] = item[1]

    explTable = pd.DataFrame(np.array(list(impacts.values())).reshape(1,-1), columns=impacts.keys()).T
    explTable = explTable.rename(columns={0: 'ImpactOnPrediction'})
    explTable['Positive/Negative'] = explTable['ImpactOnPrediction'].apply(lambda x: 'Negative' if x < 0 else 'Positive')

    return pred, explTable

#-------------------------------------------------------------------------------------------
# Streamlit app
st.title("Student's Final Grade Prediction")

# Input
inputs = []
for variable in variableExplDict:
    st.write(variableExplDict[variable])
    if variable in obj_columns:
        value = st.selectbox(variable, classes_dict[variable], key=variable)  # Create a dropdown menu
    else:
        value = st.text_input(variable, key=variable)
    inputs.append(value)
    
# Predict button
if st.button("Predict"):
    score, explantn = predict_score(inputs)
    st.write("Prediction: ", score)
    st.write("Impact on prediction:", explantn)

# Clear button functionality
if st.button("Clear"):
    st.experimental_rerun()

# !streamlit run app.py