app.py

import streamlit as st
import pandas as pd
import joblib
from preprocessing import clean_text, case_folding, tokenize, normalized_term, remove_stopwords, stem_text

# Dictionary untuk menyimpan model, vectorizer, dan jumlah fitur yang diinginkan
models = {
    'Model': {'svm': 'svm_model.pkl', 'vectorizer': 'tfidf_prabowo_final.pkl', 'features': 'prabowo_selected.pkl'},
}

# Pilihan model
selected_model = st.sidebar.selectbox('Choose Model:', list(models.keys()))

# Load model dan vectorizer untuk teks
selected_model = models[selected_model]
model = joblib.load(selected_model['svm'])
vectorizer = joblib.load(selected_model['vectorizer'])
features = joblib.load(selected_model['features'])

st.title('Website Prediksi Analisis Sentimen Calon Presiden Indonesia 2024')

# # Daftar kata-kata yang langsung diklasifikasikan sebagai Positif
positive_keywords = ['menang', 'bagus', 'keren', 'mantap', 'bijak', 'berhasil']
negative_keywords = ['bangsat', 'anjing', 'hoax','tai','babi','bajingan','penipu','pembohong','telek','sialan', 'gagal','kalah','jelek']

# Fungsi untuk memeriksa apakah teks mengandung kata-kata positif
def contains_positive_keywords(text, keywords):
    for word in keywords:
        if word in text:
            return True
    return False

# Fungsi untuk memeriksa apakah teks mengandung kata-kata positif
def contains_negative_keywords(text, keywords):
    for word in keywords:
        if word in text:
            return True
    return False

# Pilihan input: Teks atau File
option = st.sidebar.selectbox('Choose Input Option:', ['Text', 'File'])

if option == 'Text':
    # Input teks
    user_input = st.text_area('Enter Text:', '')
    cleaned_text = clean_text(user_input)
    folded_text = case_folding(cleaned_text)
    tokenized_text = tokenize(folded_text)
    normalized_text = normalized_term(tokenized_text)
    wstopword_text = remove_stopwords(normalized_text)
    stemmed_text = stem_text(wstopword_text)
    done_text = ' '.join(stemmed_text)
    if st.button('Analyze'):
        if contains_positive_keywords(user_input, positive_keywords):
            st.write(pd.DataFrame({'Step': ['Cleaning', 'Case Folding', 'Tokenization', 'Normalization', 'Stopword Removal', 'Stemming'],
                                   'Result': [cleaned_text, folded_text, tokenized_text, normalized_text, wstopword_text,
                                              stemmed_text]}))
            st.write('Sentiment: Positif')
            
        elif contains_negative_keywords(user_input, negative_keywords):
            st.write(pd.DataFrame({'Step': ['Cleaning', 'Case Folding', 'Tokenization', 'Normalization', 'Stopword Removal', 'Stemming'],
                                   'Result': [cleaned_text, folded_text, tokenized_text, normalized_text, wstopword_text,
                                              stemmed_text]}))
            st.write('Sentiment: Negatif')
            
        else:
        # Preprocessing teks
            cleaned_text = clean_text(user_input)
            folded_text = case_folding(cleaned_text)
            tokenized_text = tokenize(folded_text)
            normalized_text = normalized_term(tokenized_text)
            wstopword_text = remove_stopwords(normalized_text)
            stemmed_text = stem_text(wstopword_text)
            done_text = ' '.join(stemmed_text)
        
            # Tampilkan tahapan preprocessing
            st.subheader('Preprocessing Steps:')
            st.write(pd.DataFrame({'Step': ['Cleaning', 'Case Folding', 'Tokenization', 'Normalization', 'Stopword Removal', 'Stemming'],
                                    'Result': [cleaned_text, folded_text, tokenized_text, normalized_text, wstopword_text,
                                                stemmed_text]}))
    
            # Proses teks dan lakukan prediksi
            new_data_tfidf = vectorizer.transform([done_text]).toarray()
            new_data_selected = pd.DataFrame(new_data_tfidf, columns=vectorizer.get_feature_names_out())[features].values
    
            prediction = model.predict(new_data_selected)
            st.write('Sentiment:', prediction[0])

elif option == 'File':
    # Input file Excel/CSV
    uploaded_file = st.file_uploader('Upload Excel/CSV File:', type=['csv', 'xlsx'])
    if uploaded_file is not None:
        # Baca file dan lakukan prediksi
        df = pd.read_excel(uploaded_file) if uploaded_file.name.endswith('xlsx') else pd.read_csv(uploaded_file)

        # Preprocessing teks dalam kolom 'Text'
        df['Cleaned'] = df['text'].apply(clean_text)
        df['Case Folded'] = df['Cleaned'].apply(case_folding)
        df['Tokenized'] = df['Case Folded'].apply(tokenize)
        df['Normalized'] = df['Tokenized'].apply(normalized_term)
        df['Stopword Removal'] = df['Normalized'].apply(remove_stopwords)
        df['Stemmed'] = df['Stopword Removal'].apply(stem_text)
        df['Text Final'] = df['Stemmed'].apply(lambda words: ' '.join(words))

        # Proses teks dan lakukan prediksi
        # Jumlah fitur yang diinginkan sesuai dengan model SVM
        new_data_tfidf = vectorizer.transform(df['Text Final']).toarray()
        new_data_selected = pd.DataFrame(new_data_tfidf, columns=vectorizer.get_feature_names_out())[features].values

        predictions = []
        # for text in df['text']:
        #     if contains_positive_keywords(text, positive_keywords):
        #         predictions.append('Positif')
        #     else:
        #         new_data_selected_row = new_data_selected[df['text'].tolist().index(text)].reshape(1, -1)
        #         prediction = model.predict(new_data_selected_row)
        #         predictions.append(prediction[0])

        df['Sentiment'] = predictions

        # Tampilkan jumlah label positif dan negatif
        st.subheader('Sentiment Distribution:')
        st.write(df['Sentiment'].value_counts())

        # Tampilkan tahapan preprocessing
        st.subheader('Preprocessing Steps:')
        st.write(df)