app.py

import pandas as pd
import numpy as np
import streamlit as st
from geopy.geocoders import Nominatim
from geopy.distance import geodesic
from sklearn.feature_extraction.text import CountVectorizer
from sklearn.metrics.pairwise import cosine_similarity
import logging
import warnings
import time

logging.getLogger("urllib3").setLevel(logging.CRITICAL)
warnings.filterwarnings("ignore", category=UserWarning, module="urllib3")

geolocator = Nominatim(user_agent="job_recommendation_system")

def get_coordinates(location):
    geolocator = Nominatim(user_agent="job_recommendation_system")
    location = location.replace("-", ",")  

    if "Metropolitan Area" in location or "Region" in location:
        city_name = location.split(" ")[0]  # Ambil nama kota utama
        location = city_name

    location_obj = geolocator.geocode(location)
    if location_obj:
        return location_obj.latitude, location_obj.longitude

    print(f"Koordinat untuk {location} tidak ditemukan. Mencoba alternatif.")
    city_or_area = location.split(',')[0].strip()  # Ambil nama kota pertama
    location_obj = geolocator.geocode(city_or_area)

    if not location_obj:
        country = location.split(',')[-1].strip()  # Ambil nama negara terakhir
        location_obj = geolocator.geocode(country)

    retry_count = 0
    while not location_obj and retry_count < 5:
        print(f"Mencoba ulang untuk lokasi {location}...")
        time.sleep(2)  # Delay
        location_obj = geolocator.geocode(location)
        retry_count += 1

    if location_obj:
        return location_obj.latitude, location_obj.longitude
    else:
        print(f"Koordinat untuk {location} atau alternatif tidak dapat ditemukan.")
        return None, None

sample_data = pd.read_csv('job_data_with_coordinates.csv')

# 1. Vektorisasi skill menggunakan CountVectorizer
def vectorize_skills(skills, all_skills):
    vectorizer = CountVectorizer()
    vectorizer.fit(all_skills)
    skills_vector = vectorizer.transform(skills)
    return skills_vector

# 2. Menghitung Cosine Similarity
def calculate_cosine_similarity(user_skills_tfidf, job_skills_tfidf):
    return cosine_similarity(user_skills_tfidf, job_skills_tfidf)

# 3. Menghitung jarak lokasi
def calculate_distance(job_coords, user_coords):
    try:
        return geodesic(job_coords, user_coords).km
    except ValueError:  # Menangani kasus koordinat yang tidak valid
        return float('inf')

# Fungsi normalisasi jarak menggunakan log
def normalize_distance(distances):
    normalized = 1 / (1 + np.log1p(distances))  # log1p untuk menangani log(1 + distance)
    return normalized

# Menambahkan radius sebagai parameter
def prepare_and_recommend(df, user_skills, user_location):
    # 1. Memastikan dataset memiliki koordinat
    if 'latitude' not in df or 'longitude' not in df:
        raise ValueError("Dataset harus memiliki kolom latitude dan longitude")

    # 2. Vektorisasi skill
    all_skills = df['skills'].tolist()
    user_skills_vtr = vectorize_skills([user_skills], all_skills)
    job_skills_vtr = vectorize_skills(df['skills'], all_skills)

    # 3. Menghitung Cosine Similarity antara user dan pekerjaan
    cosine_similarities = calculate_cosine_similarity(user_skills_vtr, job_skills_vtr)
    df['cosine_similarity'] = cosine_similarities[0]

    # 4. Menghitung jarak antara lokasi pekerjaan dan lokasi user
    user_coords = get_coordinates(user_location)
    distances = []
    for _, row in df.iterrows():
        if pd.notna(row['latitude']) and pd.notna(row['longitude']):
            job_coords = (row['latitude'], row['longitude'])
            distance = calculate_distance(job_coords, user_coords)
            distances.append(distance)
        else:
            distances.append(float('inf'))
    df['distance (km)'] = distances

    # 5. Filter pekerjaan berdasarkan radius
    df = df[df['distance (km)']]

    # 6. Normalisasi jarak
    if not df.empty:
        df['normalized_distance'] = normalize_distance(df['distance (km)'])

        # 7. Menghitung skor akhir
        df['final score'] = (1.5 * df['cosine_similarity']) * (1.0 * df['normalized_distance'])
        df['final score'] = df['final score'].round(2)

        # 8. Mengurutkan pekerjaan dan memilih 5 teratas berdasarkan skor akhir
        top_jobs = df.sort_values(by='final score', ascending=False).head(5)
    else:
        top_jobs = pd.DataFrame()  # Kembalikan DataFrame kosong jika tidak ada pekerjaan dalam radius

    return top_jobs[['job_link', 'title', 'company', 'location', 'distance (km)', 'final score']]

# Streamlit app
st.title('Job Findr')
st.write('Find your job with ease.')

user_skills = st.text_input('Enter your skills (comma-separated):')
user_location = st.text_input('Job location:')

if st.button('Get Recommendations'):
    if user_skills and user_location:
        recommended_jobs = prepare_and_recommend(sample_data, user_skills, user_location)
        if recommended_jobs.empty:
            st.warning('Tidak ditemukan pekerjaan yang sesuai dengan keterampilan dan lokasi Anda.')
        elif recommended_jobs['final score'].max() < 0.02:
            st.warning('Maaf, hasil rekomendasi mungkin kurang relevan dengan keterampilan dan lokasi Anda. Silakan coba mengubah keterampilan atau lokasi Anda.')
        else:
            recommended_jobs['Rank'] = range(1, len(recommended_jobs) + 1)
            columns = ['Rank', 'title', 'job_link', 'company', 'location', 'distance (km)', 'final score']
            recommended_jobs = recommended_jobs[columns]
            recommended_jobs.reset_index(drop=True, inplace=True)
            st.dataframe(recommended_jobs.reset_index(drop=True))
    else:
        st.warning('Please enter your skills and location.')