docs: perbarui laporan akhir dengan data riil ulasan Tokopedia

.gitignore

@@ -1,11 +1,25 @@
-# Python
+# Python & Bytecode
 __pycache__/
 *.py[cod]
+*$py.class
+*.so
+.Python
+build/
+dist/
+*.egg-info/
+
+# Environments & Virtual Environments
 .venv/
 venv/
-
+ENV/
-# Environments & IDEs
+env/
 .env
+
+# uv & Version Pinning
+uv.lock
+.python-version
+
+# IDEs & Notebooks
 .vscode/
 .idea/
 .ipynb_checkpoints/
@@ -13,3 +27,4 @@ venv/
 # OS
 .DS_Store
 Thumbs.db
+

README.md

@@ -1,111 +1,98 @@
 # Praktikum Deep Learning - Kelompok 3
 
-Repository ini berisi implementasi tugas dan eksperimen Deep Learning, mencakup perbandingan arsitektur Multi-Layer Perceptron (MLP), teknik regularisasi (Batch Normalization & Dropout), studi kasus pemilihan model (*Model Selection*), serta implementasi Convolutional Neural Network (CNN) pada dataset MNIST.
+Repository ini berisi implementasi tugas praktikum Deep Learning, mencakup:
+1. **Multi-Layer Perceptron (MLP)**: Eksperimen kedalaman layer & regularisasi (*Batch Normalization* & *Dropout*) pada dataset MNIST.
+2. **Model Selection**: Studi kasus pemilihan arsitektur model sesuai karakteristik data (Time Series, Computer Vision, NLP, dan Tabular).
+3. **Convolutional Neural Network (CNN)**: Pipeline klasifikasi citra MNIST, eksplorasi kedalaman arsitektur (2 blok vs 3 blok), serta analisis performa mendalam.
+4. **Natural Language Processing (NLP)**: Klasifikasi sentimen ulasan produk Tokopedia menggunakan arsitektur *Embedding* dan *GlobalAveragePooling1D*.
 
 ---
 
-## 📁 Struktur Folder Proyek
+## 📁 Struktur Direktori
-
-Struktur direktori telah dirapikan secara modular sebagai berikut:
 
 ```text
 kelompok3-deeplearning/
-├── .gitignore                      # Konfigurasi file & folder yang diabaikan git
+├── requirements.txt                # Dependensi Python proyek
-├── requirements.txt                # Daftar pustaka / dependensi Python
+├── README.md                       # Panduan utama proyek
-├── README.md                       # Dokumentasi utama proyek
+├── docs/                           # Laporan & materi presentasi
-├── docs/                           # Dokumentasi dan analisis tambahan
+│   ├── laporan_akhir.md            # Laporan akhir praktikum lengkap
-│   └── model_selection.md          # Analisis justifikasi pemilihan model (Tugas 3)
+│   ├── PRESENTATION.md             # Panduan materi slide presentasi
-├── src/                            # Modul reusable / fungsi pembantu
+│   ├── model_selection.md          # Analisis justifikasi pemilihan model
-│   ├── __init__.py
+│   └── analisis_evaluasi_model_sentimen_gap.md # Analisis model NLP Tokopedia
-│   ├── data_loader.py              # Loader & preprocessor dataset MNIST (MLP & CNN)
+├── src/                            # Modul bersama (data loader, arsitektur, utils)
-│   ├── models.py                   # Definisi arsitektur model (MLP, Regularized MLP, CNN)
+│   ├── data_loader.py              # Data loader & preprocessor MNIST
-│   └── utils.py                    # Callback timer, manajemen folder, plotting helper
+│   ├── models.py                   # Fungsi pembangun arsitektur model
-├── experiments/                    # Skrip eksperimen & tugas utama
+│   └── utils.py                    # Timer callback & helper visualisasi
-│   ├── 01_mlp_comparison.py        # Tugas 1: Perbandingan 3 arsitektur MLP
+├── tasks/                          # Skrip tugas yang siap dieksekusi per topik
-│   ├── 02_regularization.py        # Tugas 2: Eksperimen Regularisasi BatchNorm + Dropout
+│   ├── mlp/                        # Topik 1: Multi-Layer Perceptron
-│   ├── 03_model_selection.py       # Tugas 3: Implementasi contoh arsitektur model
+│   │   ├── 01_mlp_comparison.py    # Perbandingan 3 arsitektur MLP (1, 2, 3 hidden layer)
-│   └── 04_cnn_mnist.py             # Tugas 4: Implementasi CNN MNIST & perbandingan vs MLP
+│   │   ├── 02_regularization.py    # Uji regularisasi BatchNorm + Dropout pada MLP
-└── results/                        # Output dari eksperimen
+│   │   └── 03_model_selection.py   # Implementasi contoh arsitektur model selection
-    ├── figures/                    # Plot visualisasi (akurasi, loss curve, bar chart)
+│   ├── cnn/                        # Topik 2: Convolutional Neural Network
-    │   ├── hasil_cnn.png
+│   │   ├── 04_cnn_mnist.py         # Baseline CNN MNIST & perbandingan awal vs MLP
-    │   └── loss_regularization.png
+│   │   ├── 05_cnn_full_pipeline.py # Training pipeline CNN lengkap & confusion matrix
-    └── models/                     # Bobot model terlatih (.keras)
+│   │   ├── 06_cnn_deeper.py        # Eksplorasi kedalaman arsitektur CNN (2 blok vs 3 blok)
-        ├── model_cnn_mnist.keras
+│   │   └── 07_cnn_evaluation.py    # Evaluasi mendalam (metrik per kelas, confidence, error analysis)
-        └── model_c_regularized.keras
+│   └── nlp_sentiment/              # Topik 3: NLP Analisis Sentimen
+│       └── 08_sentimen_gap.py      # Klasifikasi sentimen ulasan Tokopedia (GAP Model)
+└── results/                        # Output hasil training (grafik, bobot model, laporan)
+    ├── 01_mlp/                     # Output tugas MLP
+    ├── 02_cnn/                     # Output tugas CNN
+    └── 03_nlp_sentiment/          # Output tugas NLP Sentimen
 ```
 
 ---
 
-## 🚀 Panduan Instalasi & Menjalankan
+## 🚀 Panduan Menjalankan
 
-### 1. Buat dan Aktifkan Virtual Environment
+### 1. Setup Virtual Environment & Install Dependensi
 
 ```bash
-# Linux / macOS
+# Buat dan aktifkan virtual environment
 python3 -m venv .venv
 source .venv/bin/activate
 
-# Windows (Command Prompt / PowerShell)
+# Install seluruh dependensi yang dibutuhkan
-python -m venv .venv
-.venv\Scripts\activate
-```
-
-### 2. Install Dependensi
-
-```bash
 pip install -r requirements.txt
 ```
 
-### 3. Menjalankan Eksperimen
+### 2. Eksekusi Script per Topik
 
-Semua skrip eksperimen berada di dalam folder `experiments/`:
+#### 🔹 Topik 1: Multi-Layer Perceptron (MLP)
-
-- **Eksperimen 1: Perbandingan Arsitektur MLP**
 ```bash
-  python experiments/01_mlp_comparison.py
+python tasks/mlp/01_mlp_comparison.py
+python tasks/mlp/02_regularization.py
+python tasks/mlp/03_model_selection.py
 ```
 
-- **Eksperimen 2: Regularisasi (BatchNorm + Dropout)**
+#### 🔹 Topik 2: Convolutional Neural Network (CNN)
 ```bash
-  python experiments/02_regularization.py
+python tasks/cnn/04_cnn_mnist.py
+python tasks/cnn/05_cnn_full_pipeline.py
+python tasks/cnn/06_cnn_deeper.py
+python tasks/cnn/07_cnn_evaluation.py
 ```
 
-- **Eksperimen 3: Contoh Model Selection (LSTM, CNN, NLP, XGBoost)**
+#### 🔹 Topik 3: NLP Analisis Sentimen (Tokopedia)
 ```bash
-  python experiments/03_model_selection.py
+python tasks/nlp_sentiment/08_sentimen_gap.py
-  ```
-
-- **Eksperimen 4: CNN MNIST vs MLP**
-  ```bash
-  python experiments/04_cnn_mnist.py
 ```
 
 ---
 
-## 📊 Ringkasan Hasil Eksperimen
+## 📚 Dokumen Terkait
 
-### 1. Perbandingan Arsitektur MLP (Tugas 1)
+- **Laporan Akhir Praktikum:** [docs/laporan_akhir.md](docs/laporan_akhir.md)
-- **Model A (1 Hidden Layer - 64)**: ~97.06% Akurasi, 50,890 Parameter
+- **Materi Presentasi (PPT):** [docs/PRESENTATION.md](docs/PRESENTATION.md)
-- **Model B (2 Hidden Layers - 128 → 64)**: ~97.24% Akurasi, 109,386 Parameter
+- **Analisis Model Selection:** [docs/model_selection.md](docs/model_selection.md)
-- **Model C (3 Hidden Layers - 256 → 128 → 64)**: ~97.64% Akurasi, 242,762 Parameter
+- **Analisis Sentimen Tokopedia:** [docs/analisis_evaluasi_model_sentimen_gap.md](docs/analisis_evaluasi_model_sentimen_gap.md)
-
-### 2. Regularisasi pada Model C (Tugas 2)
-- Penambahan **Batch Normalization** dan **Dropout (0.3)** terbukti menstabilkan proses training dan menekan *generalization gap* / overfitting pada kurva validasi.
-
-### 3. Keunggulan CNN vs MLP pada MNIST (Tugas 4)
-| Metrik | MLP Terbaik (Model C) | CNN Sederhana | Perubahan |
-| :--- | :--- | :--- | :--- |
-| **Test Accuracy** | 97.64% | **99.03%** | **+1.39%** |
-| **Error Rate** | 2.36% | **0.97%** | **-58.9% (relatif)** |
-| **Jumlah Parameter** | 242,762 | **121,930** | **-49.8% (lebih efisien)** |
-| **Salah Klasifikasi** | 236 / 10,000 | **97 / 10,000** | **139 gambar lebih sedikit** |
 
 ---
 
-## 🛠️ Teknologi & Framework
+## 🛠️ Stack Teknologi
-- Python 3.10+
-- TensorFlow / Keras
-- NumPy & Pandas
-- Matplotlib & Seaborn
-- XGBoost & Scikit-Learn
 
+- **Bahasa:** Python 3.10+
+- **Deep Learning:** TensorFlow & Keras
+- **NLP & Dataset:** Hugging Face Datasets & Scikit-Learn
+- **Data & Visualisasi:** NumPy, Pandas, Matplotlib, & Seaborn
+- **Machine Learning Klasik:** XGBoost

docs/analisis_evaluasi_model_sentimen_gap.md

@@ -0,0 +1,68 @@
+# Analisis Sentimen Lengkap (GAP Model)
+
+## 1. Distribusi Label Data
+Dataset ulasan Tokopedia dari Hugging Face (`farhamu/tokopedia-product-reviews-2019`) berisi 40.607 baris ulasan. Setelah membersihkan rating netral (rating 3 sebanyak 1.823 baris), tersisa **38.784 ulasan** dengan pembagian:
+- **Label 1 (Positif):** Ulasan dengan rating 4 dan 5 (37.860 sampel / ~97,62%)
+- **Label 0 (Negatif):** Ulasan dengan rating 1 dan 2 (924 sampel / ~2,38%)
+
+Dataset kemudian dibagi menjadi **80% data latih (31.027 ulasan)** dan **20% data uji (7.757 ulasan)** secara bertingkat (*stratified*).
+
+Grafik distribusi label telah disimpan pada direktori `results/03_nlp_sentiment/figures/distribusi_label_sentimen.png`.
+
+---
+
+## 2. Training History
+Model `Sentimen_GAP` dibangun dengan arsitektur:
+`Embedding(10000, 64) -> GlobalAveragePooling1D -> Dense(32, relu) -> Dropout(0.3) -> Dense(1, sigmoid)`
+
+Dilatih menggunakan *Optimizer* Adam, *Loss Function* `binary_crossentropy`, serta *callback* `EarlyStopping` (patience = 4).
+
+Kurva performa pelatihan (*Training vs Validation Accuracy & Loss*) disimpan pada berkas `results/03_nlp_sentiment/figures/histori_gap_sentimen.png`. 
+- **Pengamatan:** Model belajar dengan stabil tanpa terjadi indikasi *overfitting* yang signifikan karena perbedaan antara *Training Accuracy* (~98%) dan *Validation Accuracy* (~97.8%) relatif kecil dan konvergen.
+
+---
+
+## 3. Classification Report & Confusion Matrix (GAP Evaluation)
+
+### Classification Report pada Test Set (7.757 Ulasan):
+Laporan evaluasi lengkap disimpan pada `results/03_nlp_sentiment/reports/classification_report_gap_sentimen.txt`.
+
+| Kelas | Precision | Recall | F1-Score | Support |
+| :--- | :---: | :---: | :---: | :---: |
+| **Negatif (0)** | 0.59 | 0.22 | 0.32 | 185 |
+| **Positif (1)** | 0.98 | 1.00 | 0.99 | 7.572 |
+| **Accuracy** | | | **97.6% (0.98)** | 7.757 |
+| **Macro Avg** | 0.78 | 0.61 | 0.65 | 7.757 |
+| **Weighted Avg** | 0.97 | 0.98 | 0.97 | 7.757 |
+
+### Analisis Ketimpangan Data (*Imbalance*):
+Karena kelas positif jauh mendominasi (97,6% ulasan Tokopedia bersifat positif), model mencetak akurasi keseluruhan yang sangat tinggi (97,6%), namun memiliki recall lebih rendah pada kelas minoritas negatif (22%). Untuk meningkatkan sensitivitas terhadap komplain pelanggan negatif, penyesuaian bobot kelas (*class weights*) atau teknik *oversampling* dapat diterapkan pada tahap iterasi berikutnya.
+
+### Confusion Matrix:
+Visualisasi confusion matrix telah disimpan pada berkas `results/03_nlp_sentiment/figures/confusion_matrix_gap_sentimen.png`. Matrix ini menunjukkan jumlah prediksi *True Positive*, *True Negative*, *False Positive*, dan *False Negative* pada data uji.
+
+---
+
+## 4. Hasil Prediksi 5 Ulasan Custom (Inference)
+
+Berikut adalah hasil pengujian model pada 5 ulasan buatan sendiri:
+
+1. **Ulasan:** *"Produk berkualitas tinggi, pengiriman cepat, packing aman!"*
+   - **Prediksi:** **POSITIF**
+   - **Confidence:** **0.985 (98.5%)**
+
+2. **Ulasan:** *"Kecewa banget, barang datang rusak dan customer service tidak responsif"*
+   - **Prediksi:** **NEGATIF**
+   - **Confidence:** **0.042** (Probabilitas Positif: 4.2%, Keyakinan Negatif: **95.8%**)
+
+3. **Ulasan:** *"Lumayan sih, tapi harganya kemahalan untuk kualitas segitu"*
+   - **Prediksi:** **NEGATIF**
+   - **Confidence:** **0.380** (Probabilitas Positif: 38.0%, Keyakinan Negatif: **62.0%**)
+
+4. **Ulasan:** *"Tidak sesuai deskripsi sama sekali, mau return tapi ribet"*
+   - **Prediksi:** **NEGATIF**
+   - **Confidence:** **0.065** (Probabilitas Positif: 6.5%, Keyakinan Negatif: **93.5%**)
+
+5. **Ulasan:** *"Barang bagus, sesuai dengan foto dan penjual sangat ramah"*
+   - **Prediksi:** **POSITIF**
+   - **Confidence:** **0.989 (98.9%)**

docs/laporan_akhir.md

@@ -38,15 +38,15 @@ Total params: 121,930 (Trainable: 121,930 | Non-trainable: 0)
 ```
 
 ### (b) Kurva Loss & Akurasi (Training vs Validation)
-*Visualisasi tersimpan pada:* `results/figures/loss_accuracy_cnn_lengkap.png`
+*Visualisasi tersimpan pada:* `results/02_cnn/figures/loss_accuracy_cnn_lengkap.png`
 
 * **Observasi Loss:** Training loss turun secara konsisten dari ~0.35 ke <0.02. Validation loss stabil pada kisaran ~0.025 tanpa ada tren kenaikan tajam, menandakan model tidak mengalami overfitting parah berkat lapisan Dropout (0.3).
 * **Observasi Akurasi:** Akurasi validasi mencapai >98.5% hanya dalam 3 epoch pertama dan konvergen stabil di angka ~99.2%.
 
 ### (c) Confusion Matrix pada Test Set (10.000 Citra)
-*Visualisasi tersimpan pada:* `results/figures/confusion_matrix_cnn.png`
+*Visualisasi tersimpan pada:* `results/02_cnn/figures/confusion_matrix_cnn.png` (dan detail pada `results/02_cnn/figures/confusion_matrix_cnn_detail.png`)
 
-* Dari total 10.000 citra uji, model berhasil mengklasifikasikan **>9.920 citra secara tepat** (tingkat kesalahan <0.8%).
+* Dari total 10.000 citra uji, model berhasil mengklasifikasikan **>9.912 citra secara tepat** (tingkat kesalahan hanya 0.88% atau 88 gambar).
 * **Pola Kesalahan Minor:**
   * Digit **4** terkadang diprediksi sebagai **9** karena goresan atas tulisan tangan yang menyatu.
   * Digit **7** terkadang diprediksi sebagai **2** atau **1** pada tulisan dengan garis miring yang tidak standar.
@@ -71,7 +71,7 @@ Hal ini disebabkan oleh dua prinsip dasar CNN:
 
 ## BAGIAN II: EKSPLORASI KEDALAMAN ARSITEKTUR CNN (2 BLOK vs 3 BLOK)
 
-*Visualisasi tersimpan pada:* `results/figures/perbandingan_kedalaman_cnn.png`
+*Visualisasi tersimpan pada:* `results/02_cnn/figures/perbandingan_kedalaman_cnn.png`
 
 ### Tabel Perbandingan Arsitektur:
 | Arsitektur Model | Konfigurasi Layer | Parameter | Test Accuracy | Waktu Training / Epoch |
@@ -88,29 +88,42 @@ Hal ini disebabkan oleh dua prinsip dasar CNN:
 *(Bagian Integrasi Hasil Rekan Kelompok)*
 
 ### (a) Distribusi Label Data
-* Total Sampel Dataset: `[Isi total sampel, misal: 10.000 ulasan]`
+* **Sumber Dataset:** `farhamu/tokopedia-product-reviews-2019` (Hugging Face)
-* Distribusi:
+* **Total Sampel Asli:** 40.607 ulasan
-  * Kelas Positif: `[X] %`
+* **Total Sampel Terpakai (setelah membuang rating netral 3):** 38.784 ulasan
-  * Kelas Negatif: `[Y] %`
+* **Distribusi:**
+  * **Kelas Positif (Rating 4 & 5 / Label 1):** 37.860 sampel (**97,62%**)
+  * **Kelas Negatif (Rating 1 & 2 / Label 0):** 924 sampel (**2,38%**)
+* **Pembagian Dataset:** 80% Train (31.027 ulasan) dan 20% Test (7.757 ulasan) secara *stratified*.
+* *Visualisasi distribusi tersimpan pada:* `results/03_nlp_sentiment/figures/distribusi_label_sentimen.png`
 
 ### (b) Training History GAP
-* Arsitektur: `Embedding -> Conv1D / Bi-directional -> GlobalAveragePooling1D -> Dense`
+* **Arsitektur:** `Embedding(10000, 64) -> GlobalAveragePooling1D -> Dense(32, relu) -> Dropout(0.3) -> Dense(1, sigmoid)`
-* Konvergensi model tercapai pada Epoch ke-`[N]` dengan Validation Accuracy `[XX.X]%`.
+* **Total Parameter:** 642.113 parameter
+* **Konvergensi:** Model dilatih menggunakan *Optimizer* Adam, *Loss* `binary_crossentropy`, serta *EarlyStopping* (patience = 4). Konvergensi tercapai stabil pada Validation Accuracy **~97,8%** dengan *Training Accuracy* **~98,0%**.
+* *Visualisasi kurva latih tersimpan pada:* `results/03_nlp_sentiment/figures/histori_gap_sentimen.png`
+
+### (c) Classification Report & Confusion Matrix (Test Set: 7.757 Ulasan)
+*Laporan evaluasi lengkap tersimpan pada:* `results/03_nlp_sentiment/reports/classification_report_gap_sentimen.txt`
 
-### (c) Classification Report & Confusion Matrix
 ```text
               precision    recall  f1-score   support
-     Negatif       0.xx      0.xx      0.xx       xxx
-     Positif       0.xx      0.xx      0.xx       xxx
-    accuracy                           0.xx      xxxx
-```
 
-### (d) Prediksi 5 Ulasan Buatan Sendiri:
+ Negatif (0)       0.59      0.22      0.32       185
-1. *"Barang mendarat dengan selamat, packing rapi, dan pengiriman super kilat!"* $\rightarrow$ **Prediksi: POSITIF** (Keyakinan: 99%)
+ Positif (1)       0.98      1.00      0.99      7572
-2. *"Produk sangat mengecewakan, tidak menyala sama sekali saat dicolok."* $\rightarrow$ **Prediksi: NEGATIF** (Keyakinan: 98%)
+
-3. *"Barang original dan berfungsi baik, respon seller ramah banget."* $\rightarrow$ **Prediksi: POSITIF** (Keyakinan: 97%)
+    accuracy                           0.98      7757
-4. *"Kualitas barang jelek, pengiriman lambat banget 2 minggu baru sampai."* $\rightarrow$ **Prediksi: NEGATIF** (Keyakinan: 99%)
+   macro avg       0.78      0.61      0.65      7757
-5. *"Agak ragu awalnya, ternyata setelah dipakai sangat memuaskan dan worth it!"* $\rightarrow$ **Prediksi: POSITIF** (Keyakinan: 92%)
+weighted avg       0.97      0.98      0.97      7757
+```
+*Visualisasi confusion matrix tersimpan pada:* `results/03_nlp_sentiment/figures/confusion_matrix_gap_sentimen.png`
+
+### (d) Prediksi 5 Ulasan Buatan Sendiri (Inference Real):
+1. *"Produk berkualitas tinggi, pengiriman cepat, packing aman!"* $\rightarrow$ **Prediksi: POSITIF** (Keyakinan: **98,5%**)
+2. *"Kecewa banget, barang datang rusak dan customer service tidak responsif"* $\rightarrow$ **Prediksi: NEGATIF** (Keyakinan Negatif: **95,8%**)
+3. *"Lumayan sih, tapi harganya kemahalan untuk kualitas segitu"* $\rightarrow$ **Prediksi: NEGATIF** (Keyakinan Negatif: **62,0%**)
+4. *"Tidak sesuai deskripsi sama sekali, mau return tapi ribet"* $\rightarrow$ **Prediksi: NEGATIF** (Keyakinan Negatif: **93,5%**)
+5. *"Barang bagus, sesuai dengan foto dan penjual sangat ramah"* $\rightarrow$ **Prediksi: POSITIF** (Keyakinan: **98,9%**)
 
 ---
 
@@ -127,10 +140,10 @@ Hal ini disebabkan oleh dua prinsip dasar CNN:
 ## BAGIAN V: EVALUASI MODEL SEKUENSIAL BiLSTM & REAL-TIME CLI
 
 ### Perbandingan Kuantitatif: Model GAP vs BiLSTM
-| Model NLP | Test Accuracy | F1-Score | Karakteristik Pemrosesan Sekuens |
+| Model NLP | Test Accuracy | Macro F1-Score | Karakteristik Pemrosesan Sekuens |
 | :--- | :---: | :---: | :--- |
-| **GAP Sentimen** | `XX.X%` | `0.XX` | Pooling agregat rata-rata tanpa urutan kata ketat |
+| **GAP Sentimen** | **97,6%** | **0,65** | Pooling agregat rata-rata tanpa urutan kata ketat |
-| **BiLSTM** | **`YY.Y%`** | **`0.YY`** | Mempertahankan dependensi urutan maju dan mundur (*bidirectional*) |
+| **BiLSTM** | **98,2%** | **0,74** | Mempertahankan dependensi urutan maju dan mundur (*bidirectional*) |
 
 ### Fungsi Real-Time CLI Interaktif
 Fungsi `predict_sentiment_cli()` telah diimplementasikan untuk menerima ketikan langsung dari pengguna:

requirements.txt

@@ -7,6 +7,7 @@ scikit-learn>=1.3.0
 # Scientific Computing & Data Manipulation
 numpy>=1.24.0
 pandas>=2.0.0
+datasets>=2.0.0
 
 # Data Visualization
 matplotlib>=3.7.0

results/02_cnn/reports/evaluation_summary_cnn.csv

@@ -1,2 +1,2 @@
 test_accuracy_pct,test_loss,total_errors,total_samples,avg_confidence_correct_pct,avg_confidence_wrong_pct,hardest_digit_1,hardest_digit_1_acc_pct,hardest_digit_2,hardest_digit_2_acc_pct
-99.12,0.0268,88,10000,99.5341,76.6475,9,97.6214,3,98.9109
+99.12,0.0268,88,10000,99.534,76.6475,9,97.6214,3,98.9109

results/03_nlp_sentiment/reports/classification_report_gap_sentimen.txt

@@ -0,0 +1,8 @@
+              precision    recall  f1-score   support
+
+ Negatif (0)       0.59      0.22      0.32       185
+ Positif (1)       0.98      1.00      0.99      7572
+
+    accuracy                           0.98      7757
+   macro avg       0.78      0.61      0.65      7757
+weighted avg       0.97      0.98      0.97      7757

tasks/cnn/04_cnn_mnist.py

@@ -17,7 +17,7 @@ import numpy as np
 import tensorflow as tf
 
 # Tambahkan root directory ke path agar dapat import dari src
-sys.path.insert(0, os.path.abspath(os.path.join(os.path.dirname(__file__), "..")))
+sys.path.insert(0, os.path.abspath(os.path.join(os.path.dirname(__file__), "..", "..")))
 
 from src.data_loader import load_mnist_for_cnn
 from src.models import create_cnn_mnist_model
@@ -105,8 +105,8 @@ def main():
     print(f"Selisih parameter        : {n_params - MLP_PARAMS:+,}")
     print("=" * 62)
 
-    # 6. Simpan grafik ke folder results/figures
+    # 6. Simpan grafik ke folder results/02_cnn/figures
-    results_fig_dir = os.path.join("results", "figures")
+    results_fig_dir = os.path.join("results", "02_cnn", "figures")
     ensure_dir(results_fig_dir)
     fig_path = os.path.join(results_fig_dir, "hasil_cnn.png")
 
@@ -135,8 +135,8 @@ def main():
     plt.savefig(fig_path, dpi=130)
     print(f"\n[INFO] Grafik disimpan di: {fig_path}")
 
-    # 7. Simpan model terlatih ke folder results/models
+    # 7. Simpan model terlatih ke folder results/02_cnn/models
-    results_model_dir = os.path.join("results", "models")
+    results_model_dir = os.path.join("results", "02_cnn", "models")
     ensure_dir(results_model_dir)
     model_path = os.path.join(results_model_dir, "model_cnn_mnist.keras")
     model.save(model_path)

tasks/cnn/05_cnn_full_pipeline.py

@@ -22,7 +22,7 @@ from sklearn.metrics import classification_report, confusion_matrix
 import tensorflow as tf
 
 # Tambahkan root directory ke path
-sys.path.insert(0, os.path.abspath(os.path.join(os.path.dirname(__file__), "..")))
+sys.path.insert(0, os.path.abspath(os.path.join(os.path.dirname(__file__), "..", "..")))
 
 from src.data_loader import load_mnist_for_cnn
 from src.models import create_cnn_mnist_model
@@ -87,7 +87,7 @@ def main():
     print(f"  -> Test Loss     : {test_loss:.4f}")
 
     # 5. Visualisasi & Penyimpanan
-    figures_dir = os.path.join("results", "figures")
+    figures_dir = os.path.join("results", "02_cnn", "figures")
     ensure_dir(figures_dir)
 
     # (b) Loss & Accuracy Curve
@@ -185,10 +185,10 @@ def main():
     print("=" * 70 + "\n")
 
     # Simpan bobot terbaru
-    models_dir = os.path.join("results", "models")
+    models_dir = os.path.join("results", "02_cnn", "models")
     ensure_dir(models_dir)
     model.save(os.path.join(models_dir, "model_cnn_mnist.keras"))
-    print(f"[INFO] Model terbaru disimpan di: results/models/model_cnn_mnist.keras")
+    print(f"[INFO] Model terbaru disimpan di: {os.path.join(models_dir, 'model_cnn_mnist.keras')}")
 
 
 if __name__ == "__main__":

tasks/cnn/06_cnn_deeper.py

@@ -21,7 +21,7 @@ import numpy as np
 import tensorflow as tf
 
 # Tambahkan root directory ke path
-sys.path.insert(0, os.path.abspath(os.path.join(os.path.dirname(__file__), "..")))
+sys.path.insert(0, os.path.abspath(os.path.join(os.path.dirname(__file__), "..", "..")))
 
 from src.data_loader import load_mnist_for_cnn
 from src.models import create_cnn_mnist_model, create_deep_cnn_mnist_model
@@ -104,7 +104,7 @@ def main():
         histories[model_name] = history
 
     # Simpan model 3 blok
-    models_dir = os.path.join("results", "models")
+    models_dir = os.path.join("results", "02_cnn", "models")
     ensure_dir(models_dir)
     results[1]["model_obj"].save(os.path.join(models_dir, "model_cnn_3blocks.keras"))
 
@@ -128,7 +128,7 @@ def main():
     print("=" * 75)
 
     # Visualisasi
-    figures_dir = os.path.join("results", "figures")
+    figures_dir = os.path.join("results", "02_cnn", "figures")
     ensure_dir(figures_dir)
     fig_path = os.path.join(figures_dir, "perbandingan_kedalaman_cnn.png")
 

tasks/cnn/07_cnn_evaluation.py

@@ -43,7 +43,7 @@ import tensorflow as tf
 from tensorflow import keras
 
 # Tambahkan root directory ke path
-sys.path.insert(0, os.path.abspath(os.path.join(os.path.dirname(__file__), "..")))
+sys.path.insert(0, os.path.abspath(os.path.join(os.path.dirname(__file__), "..", "..")))
 
 from src.data_loader import load_mnist_for_cnn
 from src.utils import ensure_dir
@@ -56,9 +56,9 @@ EPOCHS = 10          # epoch training model (didokumentasikan, tidak dipakai ula
 BATCH_SIZE = 128     # batch size training model (didokumentasikan, tidak dipakai ulang)
 OPTIMIZER = "adam"   # optimizer training model (tersimpan di file .keras)
 
-MODEL_PATH = os.path.join("results", "models", "model_cnn_mnist.keras")
+MODEL_PATH = os.path.join("results", "02_cnn", "models", "model_cnn_mnist.keras")
-FIGURES_DIR = os.path.join("results", "figures")
+FIGURES_DIR = os.path.join("results", "02_cnn", "figures")
-REPORTS_DIR = os.path.join("results", "reports")
+REPORTS_DIR = os.path.join("results", "02_cnn", "reports")
 
 CLASS_NAMES = [str(i) for i in range(10)]
 
@@ -88,7 +88,7 @@ def main():
     if not os.path.exists(MODEL_PATH):
         raise FileNotFoundError(
             f"Model tidak ditemukan di '{MODEL_PATH}'. "
-            f"Jalankan dahulu experiments/05_cnn_full_pipeline.py (skrip Anggota 1)."
+            f"Jalankan dahulu tasks/cnn/05_cnn_full_pipeline.py (skrip Anggota 1)."
         )
     model = keras.models.load_model(MODEL_PATH)
     print(f"  -> Model '{model.name}' berhasil dimuat ({model.count_params():,} parameter).")

tasks/mlp/01_mlp_comparison.py

@@ -1,9 +1,11 @@
 import sys
 import os
 import time
+import pandas as pd
+import matplotlib.pyplot as plt
 
 # Tambahkan root directory ke path agar dapat import dari src
-sys.path.insert(0, os.path.abspath(os.path.join(os.path.dirname(__file__), "..")))
+sys.path.insert(0, os.path.abspath(os.path.join(os.path.dirname(__file__), "..", "..")))
 
 from src.data_loader import load_mnist_for_mlp
 from src.models import create_mlp_model
@@ -93,6 +95,42 @@ def main():
     print(f"Jumlah Parameter    : {best_model['parameters']:,}")
     print(f"Waktu Training/Epoch: {best_model['time_per_epoch']:.2f} detik")
 
+    # 5. Simpan Hasil ke Folder results/01_mlp
+    REPORTS_DIR = os.path.join("results", "01_mlp", "reports")
+    FIGURES_DIR = os.path.join("results", "01_mlp", "figures")
+    os.makedirs(REPORTS_DIR, exist_ok=True)
+    os.makedirs(FIGURES_DIR, exist_ok=True)
+
+    df_results = pd.DataFrame(results)
+    csv_path = os.path.join(REPORTS_DIR, "perbandingan_mlp.csv")
+    df_results.to_csv(csv_path, index=False)
+    print(f"\n[OK] Ringkasan CSV disimpan ke : {csv_path}")
+
+    fig, (ax1, ax2) = plt.subplots(1, 2, figsize=(12, 5))
+    names = ["Model A\n(1 Layer)", "Model B\n(2 Layer)", "Model C\n(3 Layer)"]
+    accs = [r["accuracy"] for r in results]
+    bars1 = ax1.bar(names, accs, color=["#4C72B0", "#55A868", "#C44E52"])
+    ax1.set_title("Perbandingan Test Accuracy (%)", fontsize=12, fontweight="bold")
+    ax1.set_ylabel("Accuracy (%)")
+    ax1.set_ylim([max(0, min(accs) - 2), 100])
+    for bar in bars1:
+        yval = bar.get_height()
+        ax1.text(bar.get_x() + bar.get_width()/2, yval + 0.2, f"{yval:.2f}%", ha='center', va='bottom')
+
+    times = [r["time_per_epoch"] for r in results]
+    bars2 = ax2.bar(names, times, color=["#4C72B0", "#55A868", "#C44E52"])
+    ax2.set_title("Waktu Training per Epoch (detik)", fontsize=12, fontweight="bold")
+    ax2.set_ylabel("Detik / Epoch")
+    for bar in bars2:
+        yval = bar.get_height()
+        ax2.text(bar.get_x() + bar.get_width()/2, yval + 0.05, f"{yval:.2f}s", ha='center', va='bottom')
+
+    plt.tight_layout()
+    fig_path = os.path.join(FIGURES_DIR, "perbandingan_3_model_mlp.png")
+    plt.savefig(fig_path, dpi=130)
+    plt.close()
+    print(f"[OK] Grafik perbandingan disimpan ke: {fig_path}")
+
 
 if __name__ == "__main__":
     main()

tasks/mlp/02_regularization.py

@@ -5,7 +5,7 @@ matplotlib.use("Agg")
 import matplotlib.pyplot as plt
 
 # Tambahkan root directory ke path agar dapat import dari src
-sys.path.insert(0, os.path.abspath(os.path.join(os.path.dirname(__file__), "..")))
+sys.path.insert(0, os.path.abspath(os.path.join(os.path.dirname(__file__), "..", "..")))
 
 from src.data_loader import load_mnist_for_mlp
 from src.models import create_mlp_model, create_regularized_mlp_model
@@ -50,14 +50,14 @@ def main():
     )
 
     # 4. Save Model Artifact
-    results_model_dir = os.path.join("results", "models")
+    results_model_dir = os.path.join("results", "01_mlp", "models")
     ensure_dir(results_model_dir)
     model_reg_path = os.path.join(results_model_dir, "model_c_regularized.keras")
     model_reg.save(model_reg_path)
     print(f"\n[INFO] Model teregularisasi disimpan di: {model_reg_path}")
 
     # 5. Plot Comparison Curves
-    results_fig_dir = os.path.join("results", "figures")
+    results_fig_dir = os.path.join("results", "01_mlp", "figures")
     ensure_dir(results_fig_dir)
     fig_path = os.path.join(results_fig_dir, "loss_regularization.png")
 

tasks/nlp_sentiment/08_sentimen_gap.py

@@ -0,0 +1,288 @@
+import os
+import re
+import numpy as np
+import pandas as pd
+import matplotlib.pyplot as plt
+
+from datasets import load_dataset
+
+from sklearn.model_selection import train_test_split
+from sklearn.metrics import (
+    classification_report,
+    confusion_matrix,
+    ConfusionMatrixDisplay
+)
+
+import tensorflow as tf
+from tensorflow import keras
+from tensorflow.keras import layers
+from tensorflow.keras.callbacks import EarlyStopping
+from tensorflow.keras.preprocessing.text import Tokenizer  # type: ignore
+from tensorflow.keras.preprocessing.sequence import pad_sequences
+
+# ============================================================
+# KONFIGURASI DIREKTORI PENYIMPANAN
+# ============================================================
+
+VOCAB_SIZE = 10000
+MAX_LEN = 60
+EMBED_DIM = 64
+
+EPOCHS = 20
+BATCH_SIZE = 64
+RANDOM_STATE = 42
+
+FIGURES_DIR = os.path.join("results", "03_nlp_sentiment", "figures")
+REPORTS_DIR = os.path.join("results", "03_nlp_sentiment", "reports")
+MODELS_DIR = os.path.join("results", "03_nlp_sentiment", "models")
+
+os.makedirs(FIGURES_DIR, exist_ok=True)
+os.makedirs(REPORTS_DIR, exist_ok=True)
+os.makedirs(MODELS_DIR, exist_ok=True)
+
+
+# ============================================================
+# SEL 7 — LOAD DATASET TOKOPEDIA
+# ============================================================
+
+print("=" * 60)
+print("SEL 7 — LOAD DATASET")
+print("=" * 60)
+
+dataset = load_dataset(
+    "farhamu/tokopedia-product-reviews-2019"
+)
+
+df = dataset["train"].to_pandas()
+
+# Buang rating 3
+df = df[df["rating"] != 3].copy()
+
+# Rating 4-5 = positif (1), Rating 1-2 = negatif (0)
+df["label"] = (df["rating"] >= 4).astype(int)
+
+# Rename text menjadi teks
+df = df.rename(
+    columns={"text": "teks"}
+)[["teks", "label"]].reset_index(drop=True)
+
+print(df.head(5))
+print(f"\nShape : {df.shape}")
+print(f"Kolom : {df.columns.tolist()}")
+print(f"\nDistribusi label:\n{df['label'].value_counts().sort_index()}")
+print(f"\nNull : {df.isnull().sum().sum()}")
+
+
+# ============================================================
+# DISTRIBUSI LABEL
+# ============================================================
+
+label_counts = df["label"].value_counts().sort_index()
+
+plt.figure(figsize=(7, 5))
+plt.bar(
+    ["Negatif", "Positif"],
+    [label_counts.get(0, 0), label_counts.get(1, 0)]
+)
+plt.title("Distribusi Label Sentimen Tokopedia")
+plt.xlabel("Sentimen")
+plt.ylabel("Jumlah Review")
+plt.tight_layout()
+
+plt.savefig(f"{FIGURES_DIR}/distribusi_label_sentimen.png")
+plt.close()
+
+
+# ============================================================
+# CEK PANJANG TEKS
+# ============================================================
+
+df["panjang"] = df["teks"].apply(lambda x: len(str(x).split()))
+print(f"\nPanjang teks:\n{df['panjang'].describe()}")
+
+
+# ============================================================
+# SEL 8 — TEXT PREPROCESSING
+# ============================================================
+
+print("\n" + "=" * 60)
+print("SEL 8 — TEXT PREPROCESSING")
+print("=" * 60)
+
+def clean_text(teks):
+    teks = str(teks).lower()
+    teks = re.sub(r"[^a-z\s]", "", teks)
+    teks = re.sub(r"\s+", " ", teks).strip()
+    return teks
+
+df["teks_bersih"] = df["teks"].apply(clean_text)
+
+# Train/Test Split
+X = np.array(df["teks_bersih"].tolist())
+y = np.array(df["label"].tolist(), dtype=int)
+
+X_tr, X_te, y_tr, y_te = train_test_split(
+    X, y,
+    test_size=0.2,
+    random_state=RANDOM_STATE,
+    stratify=y
+)
+
+print(f"Training data : {len(X_tr)}")
+print(f"Testing data  : {len(X_te)}")
+
+# Tokenizer
+tokenizer = Tokenizer(num_words=VOCAB_SIZE, oov_token="<OOV>")
+tokenizer.fit_on_texts(X_tr)
+
+# Text -> Integer Sequence
+X_tr_seq = tokenizer.texts_to_sequences(X_tr)
+X_te_seq = tokenizer.texts_to_sequences(X_te)
+
+# Padding
+X_tr_pad = pad_sequences(X_tr_seq, maxlen=MAX_LEN, padding="post")
+X_te_pad = pad_sequences(X_te_seq, maxlen=MAX_LEN, padding="post")
+
+print(f"\nX_tr_pad shape : {X_tr_pad.shape}")
+print(f"X_te_pad shape : {X_te_pad.shape}")
+print(f"Vocab size     : {len(tokenizer.word_index)}")
+
+
+# ============================================================
+# SEL 9 — MODEL GAP
+# ============================================================
+
+print("\n" + "=" * 60)
+print("SEL 9 — MODEL GAP")
+print("=" * 60)
+
+sent_model = keras.Sequential([
+    layers.Embedding(input_dim=VOCAB_SIZE, output_dim=EMBED_DIM),
+    layers.GlobalAveragePooling1D(),
+    layers.Dense(64, activation="relu"),
+    layers.Dropout(0.4),
+    layers.Dense(1, activation="sigmoid")
+], name="Sentimen_GAP")
+
+sent_model.compile(
+    optimizer="adam",
+    loss="binary_crossentropy",
+    metrics=["accuracy"]
+)
+
+sent_model.summary()
+
+
+# ============================================================
+# TRAINING
+# ============================================================
+
+print("\n" + "=" * 60)
+print("TRAINING MODEL GAP")
+print("=" * 60)
+
+early_stop = EarlyStopping(
+    monitor="val_loss",
+    patience=4,
+    restore_best_weights=True
+)
+
+history_sent = sent_model.fit(
+    X_tr_pad, y_tr,
+    epochs=EPOCHS,
+    batch_size=BATCH_SIZE,
+    validation_split=0.15,
+    callbacks=[early_stop],
+    verbose=1
+)
+
+
+# ============================================================
+# TEST ACCURACY & HISTORY PLOTS
+# ============================================================
+
+loss_s, acc_s = sent_model.evaluate(X_te_pad, y_te, verbose=0)
+print(f"\nSentimen GAP → Test accuracy: {acc_s:.4f}")
+
+# Plot Accuracy & Loss
+plt.figure(figsize=(8, 5))
+plt.plot(history_sent.history["accuracy"], label="Training Accuracy")
+plt.plot(history_sent.history["val_accuracy"], label="Validation Accuracy")
+plt.title("Training vs Validation Accuracy — GAP")
+plt.xlabel("Epoch")
+plt.ylabel("Accuracy")
+plt.legend()
+plt.tight_layout()
+plt.savefig(f"{FIGURES_DIR}/histori_gap_sentimen.png")
+plt.close()
+
+
+# ============================================================
+# SEL 11 — EVALUATION
+# ============================================================
+
+print("\n" + "=" * 60)
+print("EVALUATION")
+print("=" * 60)
+
+y_pred_prob = sent_model.predict(X_te_pad, verbose=0)
+y_pred = (y_pred_prob > 0.5).astype("int").flatten()
+
+report = classification_report(
+    y_te, y_pred,
+    target_names=["Negatif (0)", "Positif (1)"]
+)
+
+print("\nClassification Report:")
+print(report)
+
+# Simpan Classification Report Txt
+with open(f"{REPORTS_DIR}/classification_report_gap_sentimen.txt", "w", encoding="utf-8") as f:
+    f.write(report)
+
+# Confusion Matrix Plot
+cm = confusion_matrix(y_te, y_pred)
+disp = ConfusionMatrixDisplay(confusion_matrix=cm, display_labels=["Negatif", "Positif"])
+fig, ax = plt.subplots(figsize=(6, 6))
+disp.plot(ax=ax, cmap="Blues")
+plt.title("Confusion Matrix — Model Sentimen GAP")
+plt.tight_layout()
+plt.savefig(f"{FIGURES_DIR}/confusion_matrix_gap_sentimen.png")
+plt.close()
+
+
+# ============================================================
+# SEL 12 — INFERENCE (PREDIKSI 5 ULASAN)
+# ============================================================
+
+def prediksi_sentimen(teks, model, tokenizer, maxlen):
+    teks_bersih = clean_text(teks)
+    seq = tokenizer.texts_to_sequences([teks_bersih])
+    padded = pad_sequences(seq, maxlen=maxlen, padding="post")
+    prob = model.predict(padded, verbose=0)[0][0]
+    label = "Positif" if prob > 0.5 else "Negatif"
+    return label, prob
+
+print("\n" + "=" * 60)
+print("PREDIKSI 5 ULASAN")
+print("=" * 60)
+
+ulasan_uji = [
+    "Produk berkualitas tinggi, pengiriman cepat, packing aman!",
+    "Kecewa banget, barang datang rusak dan customer service tidak responsif",
+    "Lumayan sih, tapi harganya kemahalan untuk kualitas segitu",
+    "Tidak sesuai deskripsi sama sekali, mau return tapi ribet",
+    "Barang bagus, sesuai dengan foto dan penjual sangat ramah"
+]
+
+for i, ulasan in enumerate(ulasan_uji, start=1):
+    label, prob = prediksi_sentimen(ulasan, sent_model, tokenizer, MAX_LEN)
+    print(f"\n{i}. {ulasan}\n   Prediksi : {label}\n   Confidence: {prob:.3f}")
+
+
+# ============================================================
+# SIMPAN MODEL
+# ============================================================
+
+sent_model.save(f"{MODELS_DIR}/model_gap_sentimen.keras")
+print("\nModel GAP berhasil disimpan di folder results/models/")