z4weth commited on
Commit
183c8ed
·
verified ·
1 Parent(s): 542f6d6

Upload 13 files

Browse files
.gitattributes CHANGED
@@ -33,3 +33,6 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
 
 
 
 
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
36
+ Dataset/english_jobs.csv filter=lfs diff=lfs merge=lfs -text
37
+ Dataset/jobs.csv filter=lfs diff=lfs merge=lfs -text
38
+ Siamese/recommender_model_final.keras filter=lfs diff=lfs merge=lfs -text
Dataset/english_jobs.csv ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8139533582da83eb2f64ebbf23bc3466fd3606052dc14a5456930add0fc30697
3
+ size 357837231
Dataset/jobs.csv ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4cb8539466e17e0bceb2222ea642f39587af43bf944701900a59926c69df82c8
3
+ size 178426678
Dockerfile ADDED
@@ -0,0 +1,15 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # Gunakan Python versi 3.10
2
+ FROM python:3.10
3
+
4
+ # Atur folder kerja di dalam server
5
+ WORKDIR /code
6
+
7
+ # Salin file requirements dan install library
8
+ COPY ./requirements.txt /code/requirements.txt
9
+ RUN pip install --no-cache-dir --upgrade -r /code/requirements.txt
10
+
11
+ # Salin semua file model dan kodemu ke dalam server
12
+ COPY . .
13
+
14
+ # Beri tahu server untuk menjalankan FastAPI di port 7860 (Syarat wajib Hugging Face)
15
+ CMD ["uvicorn", "main:app", "--host", "0.0.0.0", "--port", "7860"]
Siamese/pca_siamese_final.pkl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ad9ea8fdfc4621c9591d5cd40294d9c95be4e75f67b16f4a6b416f6cc8fc39b1
3
+ size 8484599
Siamese/recommender_model_final.keras ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cc1a2ae42d8d295ff9326c9454d1d44d4feb8ab8ebb5561067129079eff9d9ed
3
+ size 2098114
Siamese/vectorizer_siamese_final.pkl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9c7cae7e75eb04cd8733d6dd0d42578eded7121059247c7750a18611f6d6ab05
3
+ size 78255
main.py ADDED
@@ -0,0 +1,209 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ from fastapi import FastAPI, UploadFile, File
2
+ from deep_translator import GoogleTranslator
3
+ import pandas as pd
4
+ import numpy as np
5
+ import pickle
6
+ import tensorflow as tf
7
+ import PyPDF2
8
+ import io
9
+ import urllib.parse
10
+ import os
11
+
12
+ original_dense_init = tf.keras.layers.Dense.__init__
13
+
14
+ def safe_dense_init(self, *args, **kwargs):
15
+ kwargs.pop('quantization_config', None)
16
+ original_dense_init(self, *args, **kwargs)
17
+
18
+ tf.keras.layers.Dense.__init__ = safe_dense_init
19
+
20
+ app = FastAPI(title="NextStep AI API")
21
+
22
+ @app.get("/")
23
+ async def root():
24
+ return {"message": "Selamat datang di NextStep AI API! Silakan akses /docs untuk dokumentasi."}
25
+
26
+ print(" Memuat Dataset & Model...")
27
+
28
+ # ==================== LOAD DATASET ====================
29
+ df_job = pd.read_csv('Dataset/english_jobs.csv')
30
+
31
+ # Tambahan pengaman: mencetak kolom untuk melihat isi dataset aslimu
32
+ print("-> Kolom di english_jobs.csv:", df_job.columns.tolist())
33
+
34
+ try:
35
+ if os.path.exists('Dataset/jobs.csv'):
36
+ df_metadata = pd.read_csv('Dataset/jobs.csv')
37
+
38
+ # Hapus duplikat agar data tidak berlipat ganda, lalu ambil 3 kolom penting
39
+ df_metadata_clean = df_metadata.drop_duplicates(subset=['title'])[['title', 'company', 'url']]
40
+
41
+ # GABUNGKAN HANYA BERDASARKAN NAMA POSISI (Karena company_name tidak ada)
42
+ df_job = pd.merge(
43
+ left=df_job,
44
+ right=df_metadata_clean,
45
+ left_on='raw_job_title',
46
+ right_on='title',
47
+ how='left'
48
+ )
49
+ print("-> Sukses! Berhasil menggabungkan dataset dengan jobs.csv!")
50
+ except Exception as e:
51
+ print(f"-> Peringatan: Gagal menggabungkan jobs.csv: {e}")
52
+
53
+ df_job = df_job.dropna(subset=['extracted_skills', 'job_category'])
54
+ # ======================================================
55
+
56
+ # LOAD MODEL KLASIFIKASI
57
+ with open('models/tfidf_text.pkl', 'rb') as f:
58
+ tfidf_text = pickle.load(f)
59
+
60
+ with open('models/mlb.pkl', 'rb') as f:
61
+ mlb = pickle.load(f)
62
+
63
+ model_klasifikasi = tf.keras.models.load_model('models/best_model.h5')
64
+
65
+ with open('models/model_prediksi_kategori.pkl', 'rb') as f:
66
+ model_kategori = pickle.load(f)
67
+
68
+ # LOAD MODEL SIAMESE
69
+ class CosineSimilarityLayer(tf.keras.layers.Layer):
70
+ def __init__(self, **kwargs):
71
+ super(CosineSimilarityLayer, self).__init__(**kwargs)
72
+ def call(self, inputs):
73
+ cv_vec, job_vec = inputs
74
+ cv_norm = tf.nn.l2_normalize(cv_vec, axis=-1)
75
+ job_norm = tf.nn.l2_normalize(job_vec, axis=-1)
76
+ return tf.reduce_sum(cv_norm * job_norm, axis=-1, keepdims=True)
77
+
78
+ with open('Siamese/vectorizer_siamese_final.pkl', 'rb') as f:
79
+ vectorizer_siamese = pickle.load(f)
80
+ with open('Siamese/pca_siamese_final.pkl', 'rb') as f:
81
+ pca_siamese = pickle.load(f)
82
+
83
+ recommender_model = tf.keras.models.load_model(
84
+ 'Siamese/recommender_model_final.keras',
85
+ custom_objects={'CosineSimilarityLayer': CosineSimilarityLayer}
86
+ )
87
+ print("Semua Model Berhasil Dimuat!")
88
+
89
+ # HELPER FUNCTIONS
90
+ def ekstrak_teks_pdf(file_bytes):
91
+ teks_lengkap = ""
92
+ reader = PyPDF2.PdfReader(io.BytesIO(file_bytes))
93
+ for page in reader.pages:
94
+ teks_lengkap += page.extract_text() + " "
95
+ return teks_lengkap
96
+
97
+ def terjemahkan_cv(teks):
98
+ try:
99
+ if len(teks) > 4900:
100
+ teks = teks[:4900]
101
+ return GoogleTranslator(source='auto', target='en').translate(teks)
102
+ except Exception as e:
103
+ print(f"Translasi Gagal (Internet down/limit): {e}")
104
+ return teks
105
+
106
+ def cari_skill_gap(teks_cv, syarat_skill):
107
+ clean_syarat = str(syarat_skill).replace('[','').replace(']','').replace("'", "").split(',')
108
+ gap = [skill.strip() for skill in clean_syarat if skill.strip().lower() not in teks_cv.lower()]
109
+ return [g for g in gap if g]
110
+
111
+
112
+ # ENDPOINT API
113
+ @app.post("/api/predict-cv")
114
+ async def predict_cv(file: UploadFile = File(...)):
115
+ try:
116
+ file_bytes = await file.read()
117
+ teks_cv = ekstrak_teks_pdf(file_bytes)
118
+
119
+ vektor_cv_ai = tfidf_text.transform([teks_cv]).toarray()
120
+ vektor_dummy = np.zeros((1, 50))
121
+ vektor_input_klasifikasi = [vektor_cv_ai, vektor_dummy]
122
+
123
+ raw_pred = model_klasifikasi.predict(vektor_input_klasifikasi, verbose=0)
124
+ if isinstance(raw_pred, list):
125
+ raw_pred = raw_pred[0]
126
+
127
+ prediksi_prob = np.array(raw_pred).flatten()
128
+ y_pred_binary = (prediksi_prob > 0.50).astype(int)
129
+ skills_pelamar = mlb.inverse_transform(y_pred_binary.reshape(1, -1))[0]
130
+
131
+ kategori_prediksi = model_kategori.predict(vektor_cv_ai)[0]
132
+ df_filtered = df_job[df_job['job_category'] == kategori_prediksi]
133
+
134
+ if df_filtered.empty:
135
+ df_filtered = df_job
136
+
137
+ df_sample = df_filtered.sample(min(100, len(df_filtered)))
138
+
139
+ array_mentah = vectorizer_siamese.transform([teks_cv]).toarray()
140
+ vektor_input_siamese = pca_siamese.transform(array_mentah)
141
+
142
+ hasil_rekomendasi = []
143
+ for index, row in df_sample.iterrows():
144
+ teks_job = str(row['extracted_skills'])
145
+ array_job = vectorizer_siamese.transform([teks_job]).toarray()
146
+ vektor_job = pca_siamese.transform(array_job)
147
+
148
+ skor_ai = recommender_model.predict([vektor_input_siamese, vektor_job], verbose=0)[0][0]
149
+ persentase = round(float(((skor_ai + 1) / 2) * 100), 2)
150
+
151
+ posisi_loker = row.get('raw_job_title', f"Lowongan {index}")
152
+
153
+ # --- LOGIKA PENGAMBILAN PERUSAHAAN YANG LEBIH KETAT ---
154
+ perusahaan_loker = row.get('company') # Mencoba ambil dari jobs.csv
155
+
156
+ # Jika masih kosong, baru pakai Perusahaan Mitra
157
+ if pd.isna(perusahaan_loker) or str(perusahaan_loker).strip() == "":
158
+ perusahaan_loker = "Perusahaan Mitra"
159
+
160
+ # --- LOGIKA LINK APPLY JOB YANG LEBIH KETAT ---
161
+ link_asli = row.get('url')
162
+
163
+ if pd.isna(link_asli) or str(link_asli).strip() == "":
164
+ if perusahaan_loker == "Perusahaan Mitra":
165
+ kata_kunci = str(posisi_loker)
166
+ else:
167
+ kata_kunci = f"{posisi_loker} {perusahaan_loker}"
168
+
169
+ link_lamaran = f"https://www.linkedin.com/jobs/search/?keywords={urllib.parse.quote(kata_kunci)}"
170
+ else:
171
+ link_lamaran = str(link_asli)
172
+
173
+ hasil_rekomendasi.append({
174
+ "posisi": posisi_loker if pd.notna(posisi_loker) else f"Lowongan {index}",
175
+ "perusahaan": perusahaan_loker,
176
+ "industri": kategori_prediksi,
177
+ "skor_match": persentase,
178
+ "syarat_skill": teks_job,
179
+ "link_lamaran": link_lamaran
180
+ })
181
+
182
+ hasil_rekomendasi = sorted(hasil_rekomendasi, key=lambda x: x['skor_match'], reverse=True)[:3]
183
+
184
+ final_rekomendasi = []
185
+ for loker in hasil_rekomendasi:
186
+ gap = cari_skill_gap(teks_cv, loker['syarat_skill'])
187
+ loker['saran_belajar'] = gap[:5]
188
+
189
+ link_belajar = []
190
+ for skill in gap[:5]:
191
+ url_belajar = f"https://www.coursera.org/search?query={urllib.parse.quote(skill)}"
192
+ link_belajar.append(url_belajar)
193
+
194
+ loker['link_belajar'] = link_belajar
195
+
196
+ del loker['syarat_skill']
197
+ final_rekomendasi.append(loker)
198
+
199
+ return {
200
+ "status": "success",
201
+ "data_pelamar": {
202
+ "skill_terdeteksi": list(skills_pelamar)
203
+ },
204
+ "rekomendasi_utama_bidang": kategori_prediksi,
205
+ "top_3_loker": final_rekomendasi
206
+ }
207
+
208
+ except Exception as e:
209
+ return {"status": "error", "message": f"Terjadi kesalahan pada server: {str(e)}"}
models/best_model.h5 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0b7ccd35752adee2f193cee9eb725d6ae761c8d2bb855afc8369e191ea2f5a29
3
+ size 4941812
models/mlb.pkl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:783b20d004a441d70f04b0a1317a2dd2f00acf822a1b0542a654daed649575dc
3
+ size 1040
models/model_prediksi_kategori.pkl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:995f0da91f238d8c6c5aae9791aec0f921821d9e8dbe03ec86e4410ffcd26938
3
+ size 482247
models/tfidf_category.pkl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:da10fa1c60ba1782f6391004ff2d713c497e5fde8bfe3d0a603bbdadc61f8eb5
3
+ size 2696
models/tfidf_text.pkl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1fe0e8597506237294c81e0e7abd4fc8bc2b4c8bfc2ff223e59bb7e1f4053674
3
+ size 112941
requirements.txt ADDED
Binary file (210 Bytes). View file