app2.py

import streamlit as st
import pandas as pd
import numpy as np
import nltk
import string
from nltk.classify import NaiveBayesClassifier
import matplotlib.pyplot as plt
import matplotlib.colors as mcolors

import logging

logging.basicConfig(filename='app_log.log', level=logging.INFO)

# Load the dataset (assuming it's already available)
nltk.download('stopwords')
nltk.download('punkt')
# data_set = pd.read_csv("data2.csv")
# Load the dataset and remove duplicates
data_set = pd.read_csv("data2.csv").drop_duplicates(subset=['posts', 'type']).reset_index(drop=True)


# Function to tokenize the words
useless_words = nltk.corpus.stopwords.words("english") + list(string.punctuation)

def build_bag_of_words_features_filtered(words):
    words = nltk.word_tokenize(words)
    return {word: 1 for word in words if not word in useless_words}

# Function to classify Extraversion/Introversion
def ExtraversionIntroversion(input):
    # Extract posts from the dataset
    all_posts = pd.DataFrame()
    types = np.unique(np.array(data_set['type']))
    for j in types:
        temp1 = data_set[data_set['type'] == j]['posts']
        temp2 = []
        for i in temp1:
            temp2 += i.split('|||')
        temp3 = pd.Series(temp2)
        all_posts[j] = temp3

    # Features for the bag of words model
    features = []
    for j in types:
        temp1 = all_posts[j]
        temp1 = temp1.dropna()
        if 'I' in j or 'E' in j:
            features += [[(build_bag_of_words_features_filtered(i), 'introvert' if 'I' in j else 'extrovert') for i in temp1]]

    # Data for training
    split = []
    for i in range(len(features)):
        split += [len(features[i]) * 0.8]
    split = np.array(split, dtype=int)

    train = []
    for i in range(len(features)):
        train += features[i][:split[i]]

    # Training the models
    intro_extro = NaiveBayesClassifier.train(train)

    # Classify Extraversion/Introversion
    tokenize = build_bag_of_words_features_filtered(input)
    ie = intro_extro.classify(tokenize)
    ie_probs = intro_extro.prob_classify(tokenize)
    ie_confidence = ie_probs.prob(ie)

    return ie, ie_confidence

def describe_personality(personality_type):
    if personality_type == 'introvert':
        return "You are likely an introvert. Introverts often enjoy spending time alone, value deep connections with a few close friends, and may prefer quiet and reflective activities."
    elif personality_type == 'extrovert':
        return "You are likely an extrovert. Extroverts tend to be outgoing, social, and enjoy interacting with a wide range of people. They often thrive in dynamic and lively environments."

def plot_personality_distribution():
    personality_counts = data_set['type'].value_counts()
    plt.bar(personality_counts.index, personality_counts.values, color='skyblue')
    plt.xlabel('Personality Type')
    plt.ylabel('Count')
    plt.title('Distribution of Personality Types')
    plt.xticks(rotation=45)
    st.pyplot()


def showsEmpty():
        sample_introvert = "I am a very quiet person. I like to spend time alone. I enjoy reading and writing."
        sample_extrovert = "I am a very outgoing person. I like to spend time with friends. I enjoy going to parties and meeting new people."
            # show sample input if user input is empty -introvert and extrovert
            # show in text area
        st.text_area("Sample Input for Introvert", sample_introvert)
        st.text_area("Sample Input for Extrovert", sample_extrovert)
        
# Streamlit app
def main():
    st.title("Personality Predictor- Extraversion/Introversion")


    st.sidebar.header("User Input")
    user_input = st.sidebar.text_area("Enter your posts (separated by |||):")
    if user_input == None or user_input == "":
            showsEmpty()

    if st.sidebar.button("Predict Personality"):
        if user_input:
            result, confidence = ExtraversionIntroversion(user_input)
            description = describe_personality(result)
            st.success(f"Your predicted Personality is: {result} with confidence: {confidence:.2%} accuracy.")
            st.info(description)
            logging.info(f"Input: {user_input}, Predicted Personality: {result}, Confidence: {confidence:.2%}")
             # Display stream plot for confidence level, extrovert, introvert
            confidence = [confidence, 1-confidence]
            labels = ['Introvert', 'Extrovert']
            colors = ['lightcoral', 'lightskyblue']
            fig1, ax1 = plt.subplots()
            ax1.pie(confidence, labels=labels, colors=colors, autopct='%1.1f%%', startangle=90)
            ax1.axis('equal')
            st.pyplot(fig1)
            # plot_personality_distribution()
        


            
            

            
            
        else:
            st.warning("Please enter your posts to predict Extraversion/Introversion.")
    st.sidebar.write(f"Its nice to know more than 60%(confidence level) is accurate, but we can do better. Less than 60%(confidence level) is not very accurate. or unsure.")    
    st.sidebar.write(f"Training data: {len(data_set)} rows, {len(data_set.columns)} columns")
    st.sidebar.write(f"Due to limited data, the model may not be very accurate. {len(data_set)} rows is not enough to train a model. We need more data to improve the model.")
    
if __name__ == "__main__":
    main()