{ "nbformat": 4, "nbformat_minor": 0, "metadata": { "colab": { "provenance": [], "gpuType": "T4" }, "kernelspec": { "name": "python3", "display_name": "Python 3" }, "language_info": { "name": "python" }, "accelerator": "GPU" }, "cells": [ { "cell_type": "markdown", "source": [ "# PROPOSAL PROYEK KECERDASAN BUATAN (PJBL TAHAP I)\n", "## PENGEMBANGAN LEARNING AGENT UNTUK IDENTIFIKASI DAN KLASIFIKASI JENIS BATUAN PADA EKSPLORASI PERTAMBANGAN BERBASIS CNN\n", "\n", "---\n", "\n", "### Informasi Kelompok:\n", "* **Mata Kuliah**: Kecerdasan Buatan\n", "* **Program Studi / Kelas**: S1 Informatika / F5B2\n", "* **Kelompok**: Kelompok 1\n", "* **Dosen Pengampu**: Arif Rifai Dwiyanto, S.T., M.T.I.\n", "\n", "### Anggota Kelompok:\n", "1. **Aditya Saputra** (NIM: 202410715057)\n", "2. **Azis Andriansyah** (NIM: 202410715077)\n", "3. **Faris Nur Hakim** (NIM: 202410715142)\n", "---\n", "### Deskripsi Notebook:\n", "Notebook ini merupakan implementasi *Proof of Concept* (PoC) untuk **Soal 4 (Bab V)**[cite: 2, 293]. Sistem dirancang sebagai *Learning Agent* yang mengombinasikan *Convolutional Neural Network* (CNN) untuk klasifikasi visual citra batuan dan *decision-making engine* berbasis ambang keyakinan (*confidence thresholding*) untuk menerbitkan rekomendasi analisis laboratorium lanjutan (XRF/XRD)[cite: 9, 298].\n", "\n", "## PENTING: Pengaturan Akselerasi Perangkat Keras (T4 GPU)\n", "\n", "Sebelum mengeksekusi kode program di Google Colab, **wajib mengaktifkan akselerator GPU** agar proses pelatihan model tidak memakan waktu lama[cite: 305].\n", "\n", "### Langkah Mengaktifkan GPU:\n", "1. Klik menu **Runtime** pada bilah menu bagian atas Google Colab.\n", "2. Pilih opsi **Change runtime type** (*Ubah jenis runtime*).\n", "3. Pada bagian **Hardware accelerator** (*Akselerator perangkat keras*), pilih **T4 GPU**[cite: 305].\n", "4. Klik **Save** (*Simpan*).\n", "\n", "> **Catatan Perbandingan Kecepatan Komputasi:**\n", "> * **Menggunakan CPU bawaan:** Kecepatan pelatihan lambat $\\pm$ `6s/step` (memerlukan waktu $\\pm$ 11 menit per *epoch*)[cite: 316].\n", "> * **Menggunakan T4 GPU:** Kecepatan komputasi meningkat drastis menjadi `56ms - 126ms/step` (hanya membutuhkan $\\pm$ 7 - 13 detik per *epoch*)[cite: 305, 317, 322].\n", "\n", "---\n", "\n", "## Urutan Eksekusi Sel Notebook\n", "\n", "Eksekusi sel kode secara berurutan dari atas ke bawah (*Run All* atau jalankan sel per sel):\n", "\n", "1. **Sel Setup & Verifikasi GPU:** Memasang pustaka `kagglehub` dan memverifikasi ketersediaan akselerasi perangkat keras GPU[cite: 305].\n", "2. **Sel Unduh Dataset:** Mengunduh dan mendeteksi path direktori citra batuan secara otomatis dari repositori Kaggle (`stealthtechnologies/rock-classification`)[cite: 297].\n", "3. **Sel Pipeline Data:** Memuat dan membagi dataset ke dalam subset *Train* (80%) dan *Validation* (20%) serta mengaktifkan optimasi memori `cache()` dan `prefetch()`[cite: 167].\n", "4. **Sel Konstruksi Model:** Membangun lapisan augmentasi dan arsitektur CNN berbasis *Transfer Learning* (MobileNetV2)[cite: 10].\n", "5. **Sel Pelatihan Model:** Melatih model dengan pengawasan callback `EarlyStopping` (patience=4) yang akan berhenti otomatis saat mencapai konvergensi optimal[cite: 15, 317].\n", "6. **Sel Evaluasi & Inferensi Agen:** Menampilkan kurva akurasi/loss serta mendemonstrasikan pengujian inferensi pada sampel citra batuan untuk memvalidasi *Confidence Score* dan teks rekomendasi uji laboratorium lanjutan (XRF/XRD)[cite: 9, 293].\n", "\n", "---\n", "\n", "## Berkas Luaran yang Dihasilkan\n", "* `model_learning_agent_rock.keras`: Berkas biner model CNN terlatih yang siap digunakan pada backend aplikasi web (*Flask*) pada implementasi Tahap II[cite: 305, 310, 311, 317]." ], "metadata": { "id": "jX6B9uPwcWQw" } }, { "cell_type": "code", "execution_count": null, "metadata": { "colab": { "base_uri": "https://localhost:8080/" }, "id": "JmZLSWtgbspA", "outputId": "00b72e51-4335-4360-c6ed-4a7f991f8569" }, "outputs": [ { "output_type": "stream", "name": "stdout", "text": [ "Versi TensorFlow : 2.21.0\n", "Perangkat GPU : [PhysicalDevice(name='/physical_device:GPU:0', device_type='GPU')]\n" ] } ], "source": [ "# Instalasi pustaka pendukung dan verifikasi lingkungan akselerasi GPU\n", "!pip install -q kagglehub\n", "\n", "import os\n", "import kagglehub\n", "import numpy as np\n", "import tensorflow as tf\n", "from tensorflow.keras import layers, models\n", "import matplotlib.pyplot as plt\n", "\n", "# Verifikasi ketersediaan GPU\n", "gpu_devices = tf.config.list_physical_devices('GPU')\n", "print(\"Versi TensorFlow :\", tf.__version__)\n", "print(\"Perangkat GPU :\", gpu_devices if gpu_devices else \"GPU tidak terdeteksi (Running on CPU)\")" ] }, { "cell_type": "markdown", "source": [ "---\n", "## 1. Pengunduhan Dataset (*Kagglehub Pipeline*)\n", "Dataset diunduh langsung dari repositori publik Kaggle `stealthtechnologies/rock-classification` menggunakan pustaka `kagglehub`. Sistem secara otomatis mengidentifikasi direktori target yang memuat folder dari masing-masing kelas batuan[cite: 297]." ], "metadata": { "id": "DIw_M5iCcViA" } }, { "cell_type": "code", "source": [ "# Unduh dataset dari repositori Kaggle\n", "print(\"Mengunduh dataset dari Kaggle...\")\n", "DATASET_PATH = kagglehub.dataset_download(\"stealthtechnologies/rock-classification\")\n", "print(\"Path unduhan awal:\", DATASET_PATH)\n", "\n", "# Deteksi sub-direktori internal secara otomatis\n", "subfolders = [f.path for f in os.scandir(DATASET_PATH) if f.is_dir()]\n", "if len(subfolders) == 1:\n", " DATASET_PATH = subfolders[0]\n", "\n", "print(\"Path efektif direktori citra:\", DATASET_PATH)" ], "metadata": { "colab": { "base_uri": "https://localhost:8080/" }, "id": "1iZtGsVmcpc4", "outputId": "4d8449c2-2e5c-41d2-c3a6-dd21ad181e47" }, "execution_count": null, "outputs": [ { "output_type": "stream", "name": "stdout", "text": [ "Mengunduh dataset dari Kaggle...\n", "Using Colab cache for faster access to the 'rock-classification' dataset.\n", "Path unduhan awal: /kaggle/input/rock-classification\n", "Path efektif direktori citra: /kaggle/input/rock-classification/Rock Data\n" ] } ] }, { "cell_type": "markdown", "source": [ "---\n", "## 2. Konfigurasi Hiperparameter & Pipeline Pemuatan Data\n", "Dataset dimuat dan dibagi secara otomatis ke dalam:\n", "* **Training Set**: 80%\n", "* **Validation Set**: 20%\n", "\n", "Pipeline dioptimalkan menggunakan teknik **Caching** dan **Prefetching (AUTOTUNE)** untuk menjaga throughput GPU tetap stabil selama proses pelatihan." ], "metadata": { "id": "gbwbLhLPcuvw" } }, { "cell_type": "code", "source": [ "# ------------------------------------------------------------------------------\n", "# 2. Konfigurasi Hiperparameter & Pipeline Data (YANG DIPERBAIKI)\n", "# ------------------------------------------------------------------------------\n", "IMG_SIZE = (224, 224)\n", "BATCH_SIZE = 32\n", "EPOCHS = 20\n", "LEARNING_RATE = 1e-4\n", "\n", "# Tentukan path langsung ke sub-folder train dan valid\n", "TRAIN_DIR = os.path.join(DATASET_PATH, \"train\")\n", "VAL_DIR = os.path.join(DATASET_PATH, \"valid\")\n", "\n", "# Jika folder valid bernama 'val', deteksi otomatis\n", "if not os.path.exists(VAL_DIR):\n", " VAL_DIR = os.path.join(DATASET_PATH, \"val\")\n", "\n", "print(\"Path Data Train :\", TRAIN_DIR)\n", "print(\"Path Data Valid :\", VAL_DIR)\n", "\n", "# Memuat Dataset Train & Validation langsung dari sub-foldernya\n", "train_ds = tf.keras.utils.image_dataset_from_directory(\n", " TRAIN_DIR,\n", " image_size=IMG_SIZE,\n", " batch_size=BATCH_SIZE,\n", " seed=123\n", ")\n", "\n", "val_ds = tf.keras.utils.image_dataset_from_directory(\n", " VAL_DIR,\n", " image_size=IMG_SIZE,\n", " batch_size=BATCH_SIZE,\n", " seed=123\n", ")\n", "\n", "# Deteksi 9 Nama Kelas Batuan Aslinya\n", "CLASS_NAMES = train_ds.class_names\n", "NUM_CLASSES = len(CLASS_NAMES)\n", "print(f\"\\nJumlah Kelas Terdeteksi ({NUM_CLASSES}): {CLASS_NAMES}\")\n", "\n", "# Optimasi Performa Loading Data (Autotune & Prefetch)\n", "AUTOTUNE = tf.data.AUTOTUNE\n", "train_ds = train_ds.cache().shuffle(1000).prefetch(buffer_size=AUTOTUNE)\n", "val_ds = val_ds.cache().prefetch(buffer_size=AUTOTUNE)" ], "metadata": { "colab": { "base_uri": "https://localhost:8080/" }, "id": "6QjXRrXRc2Rf", "outputId": "c12c2a5d-cb2e-4e8e-ff40-a3960ee80899" }, "execution_count": null, "outputs": [ { "output_type": "stream", "name": "stdout", "text": [ "Path Data Train : /kaggle/input/rock-classification/Rock Data/train\n", "Path Data Valid : /kaggle/input/rock-classification/Rock Data/valid\n", "Found 3687 files belonging to 9 classes.\n", "Found 351 files belonging to 9 classes.\n", "\n", "Jumlah Kelas Terdeteksi (9): ['Basalt', 'Clay', 'Conglomerate', 'Diatomite', 'Shale-(Mudstone)', 'Siliceous-sinter', 'chert', 'gypsum', 'olivine-basalt']\n" ] } ] }, { "cell_type": "markdown", "source": [ "---\n", "## 3. Perancangan Model CNN & Augmentasi Dinamis\n", "Arsitektur model dirancang bertingkat dengan komponen:\n", "1. **Lapisan Masukan & Prapemrosesan**: `layers.Rescaling(1./255)` untuk normalisasi piksel $.\n", "2. **Augmentasi Internal**: `RandomFlip`, `RandomRotation`, dan `RandomZoom` untuk memperkaya variasi citra dan menekan *overfitting*.\n", "3. **Ekstraksi Fitur Spasial**: 3 blok `Conv2D` ($3 \\times 3$) yang dipadukan dengan `BatchNormalization`, aktivasi `ReLU`, dan `MaxPooling2D`.\n", "4. **Klasifikasi Head**: `GlobalAveragePooling2D`, `Dense` (128 unit, ReLU), regularisasi `Dropout(0.4)`, serta output `Dense` (Softmax)." ], "metadata": { "id": "eTsLLrywc6C0" } }, { "cell_type": "code", "source": [ "# ------------------------------------------------------------------------------\n", "# 3. Layer Augmentasi Data (Terintegrasi di dalam komputasi graf)\n", "# ------------------------------------------------------------------------------\n", "data_augmentation = tf.keras.Sequential([\n", " layers.RandomFlip(\"horizontal\"),\n", " layers.RandomRotation(0.1),\n", " layers.RandomZoom(0.1),\n", "], name=\"data_augmentation\")\n", "\n", "# ------------------------------------------------------------------------------\n", "# 4. Pembangunan Arsitektur CNN (TRANSFER LEARNING MOBILENETV2)\n", "# ------------------------------------------------------------------------------\n", "# 1. Gunakan Base Model MobileNetV2 yang sudah dilatih di ImageNet\n", "base_model = tf.keras.applications.MobileNetV2(\n", " input_shape=(224, 224, 3),\n", " include_top=False,\n", " weights='imagenet'\n", ")\n", "base_model.trainable = False # Bekukan bobot fitur dasar agar pelatihan cepat\n", "\n", "# 2. Rancang Arsitektur Agen\n", "model = models.Sequential([\n", " layers.Input(shape=(224, 224, 3)),\n", " layers.Rescaling(1./255),\n", " data_augmentation,\n", " base_model,\n", " layers.GlobalAveragePooling2D(),\n", " layers.Dense(128, activation='relu'),\n", " layers.Dropout(0.3),\n", " layers.Dense(NUM_CLASSES, activation='softmax')\n", "], name=\"Transfer_Learning_Rock_Classifier\")\n", "\n", "model.compile(\n", " optimizer=tf.keras.optimizers.Adam(learning_rate=1e-3),\n", " loss='sparse_categorical_crossentropy',\n", " metrics=['accuracy']\n", ")\n", "\n", "model.summary()" ], "metadata": { "colab": { "base_uri": "https://localhost:8080/", "height": 440 }, "id": "kuGCZu6FdFMM", "outputId": "7f1bdcdd-a1e4-4e1c-95b5-a72dee8dfbf4" }, "execution_count": null, "outputs": [ { "output_type": "stream", "name": "stdout", "text": [ "Downloading data from https://storage.googleapis.com/tensorflow/keras-applications/mobilenet_v2/mobilenet_v2_weights_tf_dim_ordering_tf_kernels_1.0_224_no_top.h5\n", "\u001b[1m9406464/9406464\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 0us/step\n" ] }, { "output_type": "display_data", "data": { "text/plain": [ "\u001b[1mModel: \"Transfer_Learning_Rock_Classifier\"\u001b[0m\n" ], "text/html": [ "
Model: \"Transfer_Learning_Rock_Classifier\"\n",
"\n"
]
},
"metadata": {}
},
{
"output_type": "display_data",
"data": {
"text/plain": [
"┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓\n",
"┃\u001b[1m \u001b[0m\u001b[1mLayer (type) \u001b[0m\u001b[1m \u001b[0m┃\u001b[1m \u001b[0m\u001b[1mOutput Shape \u001b[0m\u001b[1m \u001b[0m┃\u001b[1m \u001b[0m\u001b[1m Param #\u001b[0m\u001b[1m \u001b[0m┃\n",
"┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩\n",
"│ rescaling_2 (\u001b[38;5;33mRescaling\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m224\u001b[0m, \u001b[38;5;34m224\u001b[0m, \u001b[38;5;34m3\u001b[0m) │ \u001b[38;5;34m0\u001b[0m │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ data_augmentation (\u001b[38;5;33mSequential\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m224\u001b[0m, \u001b[38;5;34m224\u001b[0m, \u001b[38;5;34m3\u001b[0m) │ \u001b[38;5;34m0\u001b[0m │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ mobilenetv2_1.00_224 │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m7\u001b[0m, \u001b[38;5;34m7\u001b[0m, \u001b[38;5;34m1280\u001b[0m) │ \u001b[38;5;34m2,257,984\u001b[0m │\n",
"│ (\u001b[38;5;33mFunctional\u001b[0m) │ │ │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ global_average_pooling2d_2 │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m1280\u001b[0m) │ \u001b[38;5;34m0\u001b[0m │\n",
"│ (\u001b[38;5;33mGlobalAveragePooling2D\u001b[0m) │ │ │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ dense_4 (\u001b[38;5;33mDense\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m128\u001b[0m) │ \u001b[38;5;34m163,968\u001b[0m │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ dropout_2 (\u001b[38;5;33mDropout\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m128\u001b[0m) │ \u001b[38;5;34m0\u001b[0m │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ dense_5 (\u001b[38;5;33mDense\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m9\u001b[0m) │ \u001b[38;5;34m1,161\u001b[0m │\n",
"└─────────────────────────────────┴────────────────────────┴───────────────┘\n"
],
"text/html": [
"┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓\n",
"┃ Layer (type) ┃ Output Shape ┃ Param # ┃\n",
"┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩\n",
"│ rescaling_2 (Rescaling) │ (None, 224, 224, 3) │ 0 │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ data_augmentation (Sequential) │ (None, 224, 224, 3) │ 0 │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ mobilenetv2_1.00_224 │ (None, 7, 7, 1280) │ 2,257,984 │\n",
"│ (Functional) │ │ │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ global_average_pooling2d_2 │ (None, 1280) │ 0 │\n",
"│ (GlobalAveragePooling2D) │ │ │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ dense_4 (Dense) │ (None, 128) │ 163,968 │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ dropout_2 (Dropout) │ (None, 128) │ 0 │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ dense_5 (Dense) │ (None, 9) │ 1,161 │\n",
"└─────────────────────────────────┴────────────────────────┴───────────────┘\n",
"\n"
]
},
"metadata": {}
},
{
"output_type": "display_data",
"data": {
"text/plain": [
"\u001b[1m Total params: \u001b[0m\u001b[38;5;34m2,423,113\u001b[0m (9.24 MB)\n"
],
"text/html": [
"Total params: 2,423,113 (9.24 MB)\n", "\n" ] }, "metadata": {} }, { "output_type": "display_data", "data": { "text/plain": [ "\u001b[1m Trainable params: \u001b[0m\u001b[38;5;34m165,129\u001b[0m (645.04 KB)\n" ], "text/html": [ "
Trainable params: 165,129 (645.04 KB)\n", "\n" ] }, "metadata": {} }, { "output_type": "display_data", "data": { "text/plain": [ "\u001b[1m Non-trainable params: \u001b[0m\u001b[38;5;34m2,257,984\u001b[0m (8.61 MB)\n" ], "text/html": [ "
Non-trainable params: 2,257,984 (8.61 MB)\n", "\n" ] }, "metadata": {} } ] }, { "cell_type": "markdown", "source": [ "---\n", "## 4. Pelatihan Model dengan Early Stopping\n", "Pelatihan dipantau menggunakan callback `EarlyStopping` berbasis metrik `val_loss` dengan parameter `patience=4`. Apabila model tidak menunjukkan perbaikan generalisasi selama 4 epoch berturut-turut, pelatihan dihentikan otomatis dan bobot terbaik dipulihkan (`restore_best_weights=True`)." ], "metadata": { "id": "gpOpVK-XdIY0" } }, { "cell_type": "code", "source": [ "# Konfigurasi Callback Early Stopping\n", "callback_early_stop = tf.keras.callbacks.EarlyStopping(\n", " monitor='val_loss',\n", " patience=4,\n", " restore_best_weights=True,\n", " verbose=1\n", ")\n", "\n", "# Eksekusi Proses Pelatihan\n", "history = model.fit(\n", " train_ds,\n", " validation_data=val_ds,\n", " epochs=EPOCHS,\n", " callbacks=[callback_early_stop]\n", ")\n", "\n", "# Penyimpanan Model Format Baku Keras (.keras)\n", "MODEL_SAVE_PATH = 'model_learning_agent_rock.keras'\n", "model.save(MODEL_SAVE_PATH)\n", "print(f\"\\nModel berhasil diekspor ke: '{MODEL_SAVE_PATH}'\")" ], "metadata": { "colab": { "base_uri": "https://localhost:8080/" }, "id": "vz075dHGdM74", "outputId": "90148bfe-2c2a-4b13-a656-b28e6a9aa41b" }, "execution_count": null, "outputs": [ { "output_type": "stream", "name": "stdout", "text": [ "Epoch 1/20\n", "\u001b[1m116/116\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m26s\u001b[0m 99ms/step - accuracy: 0.4185 - loss: 1.6358 - val_accuracy: 0.5499 - val_loss: 1.2949\n", "Epoch 2/20\n", "\u001b[1m116/116\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m7s\u001b[0m 56ms/step - accuracy: 0.5829 - loss: 1.2131 - val_accuracy: 0.5385 - val_loss: 1.2869\n", "Epoch 3/20\n", "\u001b[1m116/116\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m7s\u001b[0m 59ms/step - accuracy: 0.6385 - loss: 1.0430 - val_accuracy: 0.5840 - val_loss: 1.2559\n", "Epoch 4/20\n", "\u001b[1m116/116\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m7s\u001b[0m 56ms/step - accuracy: 0.6604 - loss: 0.9759 - val_accuracy: 0.5783 - val_loss: 1.2601\n", "Epoch 5/20\n", "\u001b[1m116/116\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m7s\u001b[0m 58ms/step - accuracy: 0.6913 - loss: 0.8975 - val_accuracy: 0.6154 - val_loss: 1.2622\n", "Epoch 6/20\n", "\u001b[1m116/116\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m7s\u001b[0m 58ms/step - accuracy: 0.7234 - loss: 0.8104 - val_accuracy: 0.6182 - val_loss: 1.2238\n", "Epoch 7/20\n", "\u001b[1m116/116\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m7s\u001b[0m 60ms/step - accuracy: 0.7404 - loss: 0.7572 - val_accuracy: 0.6239 - val_loss: 1.2749\n", "Epoch 8/20\n", "\u001b[1m116/116\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m7s\u001b[0m 59ms/step - accuracy: 0.7537 - loss: 0.7141 - val_accuracy: 0.6068 - val_loss: 1.2572\n", "Epoch 9/20\n", "\u001b[1m116/116\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m7s\u001b[0m 60ms/step - accuracy: 0.7667 - loss: 0.6689 - val_accuracy: 0.6182 - val_loss: 1.3218\n", "Epoch 10/20\n", "\u001b[1m116/116\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m7s\u001b[0m 60ms/step - accuracy: 0.7817 - loss: 0.6346 - val_accuracy: 0.5954 - val_loss: 1.4003\n", "Epoch 10: early stopping\n", "Restoring model weights from the end of the best epoch: 6.\n", "\n", "Model berhasil diekspor ke: 'model_learning_agent_rock.keras'\n" ] } ] }, { "cell_type": "markdown", "source": [ "---\n", "## 5. Visualisasi Hasil Pelatihan (Subbab 5.6)\n", "Grafik perbandingan *Akurasi* dan *Loss* antara data latih dan validasi digunakan untuk mengukur ada/tidaknya celah generalisasi (*generalization gap*) dan membuktikan ketahanan model terhadap gejala *overfitting*." ], "metadata": { "id": "1-zOGWeAdPi0" } }, { "cell_type": "code", "source": [ "# Plot Evaluasi Kurva Akurasi dan Loss\n", "acc = history.history['accuracy']\n", "val_acc = history.history['val_accuracy']\n", "loss = history.history['loss']\n", "val_loss = history.history['val_loss']\n", "epochs_range = range(1, len(acc) + 1)\n", "\n", "plt.figure(figsize=(14, 5))\n", "\n", "# Kurva Akurasi\n", "plt.subplot(1, 2, 1)\n", "plt.plot(epochs_range, acc, label='Training Accuracy', marker='o')\n", "plt.plot(epochs_range, val_acc, label='Validation Accuracy', marker='s')\n", "plt.title('Kurva Akurasi Pelatihan vs Validasi')\n", "plt.xlabel('Epoch')\n", "plt.ylabel('Akurasi')\n", "plt.grid(True, linestyle='--', alpha=0.6)\n", "plt.legend(loc='lower right')\n", "\n", "# Kurva Loss\n", "plt.subplot(1, 2, 2)\n", "plt.plot(epochs_range, loss, label='Training Loss', marker='o')\n", "plt.plot(epochs_range, val_loss, label='Validation Loss', marker='s')\n", "plt.title('Kurva Loss Pelatihan vs Validasi')\n", "plt.xlabel('Epoch')\n", "plt.ylabel('Loss')\n", "plt.grid(True, linestyle='--', alpha=0.6)\n", "plt.legend(loc='upper right')\n", "\n", "plt.tight_layout()\n", "plt.show()" ], "metadata": { "id": "TR2wtT82dS48", "colab": { "base_uri": "https://localhost:8080/", "height": 376 }, "outputId": "d08d6f56-b869-48ec-f233-2cb8953930be" }, "execution_count": null, "outputs": [ { "output_type": "display_data", "data": { "text/plain": [ "