{"metadata":{"kernelspec":{"language":"python","display_name":"Python 3","name":"python3"},"language_info":{"name":"python","version":"3.7.12","mimetype":"text/x-python","codemirror_mode":{"name":"ipython","version":3},"pygments_lexer":"ipython3","nbconvert_exporter":"python","file_extension":".py"},"kaggle":{"accelerator":"nvidiaTeslaT4","dataSources":[{"sourceType":"competition","sourceId":3004,"databundleVersionId":861823}],"dockerImageVersionId":30260,"isInternetEnabled":true,"language":"python","sourceType":"notebook","isGpuEnabled":true}},"nbformat_minor":4,"nbformat":4,"cells":[{"cell_type":"code","source":"# This Python 3 environment comes with many helpful analytics libraries installed\n# It is defined by the kaggle/python Docker image: https://github.com/kaggle/docker-python\n# For example, here's several helpful packages to load\n\nimport numpy as np # linear algebra\nimport pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)\n\n# Input data files are available in the read-only \"../input/\" directory\n# For example, running this (by clicking run or pressing Shift+Enter) will list all files under the input directory\n\nimport os\nfor dirname, _, filenames in os.walk('/kaggle/input'):\n    for filename in filenames:\n        print(os.path.join(dirname, filename))\n\n# You can write up to 20GB to the current directory (/kaggle/working/) that gets preserved as output when you create a version using \"Save & Run All\" \n# You can also write temporary files to /kaggle/temp/, but they won't be saved outside of the current session","metadata":{"_uuid":"8f2839f25d086af736a60e9eeb907d3b93b6e0e5","_cell_guid":"b1076dfc-b9ad-4769-8c92-a6c4dae69d19","execution":{"iopub.status.busy":"2026-03-12T14:45:22.319065Z","iopub.execute_input":"2026-03-12T14:45:22.319458Z","iopub.status.idle":"2026-03-12T14:45:22.346965Z","shell.execute_reply.started":"2026-03-12T14:45:22.319348Z","shell.execute_reply":"2026-03-12T14:45:22.346047Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"mnist_train = pd.read_csv(\"../input/digit-recognizer/train.csv\")\nmnist_test = pd.read_csv(\"../input/digit-recognizer/test.csv\")","metadata":{"execution":{"iopub.status.busy":"2026-03-12T14:45:22.348702Z","iopub.execute_input":"2026-03-12T14:45:22.348941Z","iopub.status.idle":"2026-03-12T14:45:26.267958Z","shell.execute_reply.started":"2026-03-12T14:45:22.348919Z","shell.execute_reply":"2026-03-12T14:45:26.267160Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"y_train1 = pd.get_dummies(mnist_train.label)\nprint(y_train1.shape)","metadata":{"execution":{"iopub.status.busy":"2026-03-12T14:45:26.268932Z","iopub.execute_input":"2026-03-12T14:45:26.269221Z","iopub.status.idle":"2026-03-12T14:45:26.284916Z","shell.execute_reply.started":"2026-03-12T14:45:26.269197Z","shell.execute_reply":"2026-03-12T14:45:26.284060Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"npx = mnist_train.drop(['label'], axis=1)\nx_train1 = np.array(npx).reshape(42000,28,28,1)/255\nprint(x_train1.shape)","metadata":{"execution":{"iopub.status.busy":"2026-03-12T14:45:26.286116Z","iopub.execute_input":"2026-03-12T14:45:26.286764Z","iopub.status.idle":"2026-03-12T14:45:26.568981Z","shell.execute_reply.started":"2026-03-12T14:45:26.286735Z","shell.execute_reply":"2026-03-12T14:45:26.567941Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"from sklearn.model_selection import train_test_split #to split our dataset\nx_train, x_val, y_train, y_val = train_test_split(x_train1, y_train1, test_size=0.1, random_state=101)","metadata":{"execution":{"iopub.status.busy":"2026-03-12T14:45:26.571100Z","iopub.execute_input":"2026-03-12T14:45:26.571411Z","iopub.status.idle":"2026-03-12T14:45:27.403391Z","shell.execute_reply.started":"2026-03-12T14:45:26.571384Z","shell.execute_reply":"2026-03-12T14:45:27.402596Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"npx = mnist_test\nx_test = np.array(npx).reshape(28000,28,28,1)/255\nprint(x_test.shape)","metadata":{"execution":{"iopub.status.busy":"2026-03-12T14:45:27.404532Z","iopub.execute_input":"2026-03-12T14:45:27.404798Z","iopub.status.idle":"2026-03-12T14:45:27.544983Z","shell.execute_reply.started":"2026-03-12T14:45:27.404774Z","shell.execute_reply":"2026-03-12T14:45:27.543957Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"import keras\nfrom keras.models import Sequential\nfrom keras.optimizers import adam_v2\nfrom keras.utils.np_utils import to_categorical\nfrom keras.layers import Dense, Dropout, Flatten, BatchNormalization\nfrom keras.layers.convolutional import Conv2D, MaxPooling2D\nfrom keras.preprocessing import image\nfrom keras.callbacks import ReduceLROnPlateau\n\nimport matplotlib.pyplot as plt\nimport cv2 \nfrom PIL import Image \n\nnp.random.seed(99)\n\ndef display(img):\n  plt.figure(figsize=(5,5))\n  plt.axis(\"off\")\n  plt.imshow(img,cmap=\"gray\")\n\n\ndisplay(x_train[1191])","metadata":{"execution":{"iopub.status.busy":"2026-03-12T14:45:27.546367Z","iopub.execute_input":"2026-03-12T14:45:27.546681Z","iopub.status.idle":"2026-03-12T14:45:32.397654Z","shell.execute_reply.started":"2026-03-12T14:45:27.546653Z","shell.execute_reply":"2026-03-12T14:45:32.396073Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"def display(img):\n  plt.figure(figsize=(5,5))\n  plt.axis(\"off\")\n  plt.imshow(img,cmap=\"gray\")\n\ndisplay(x_train[1191])","metadata":{"execution":{"iopub.status.busy":"2026-03-12T14:45:32.399567Z","iopub.execute_input":"2026-03-12T14:45:32.400443Z","iopub.status.idle":"2026-03-12T14:45:32.519497Z","shell.execute_reply.started":"2026-03-12T14:45:32.400389Z","shell.execute_reply":"2026-03-12T14:45:32.517949Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"from keras.preprocessing.image import ImageDataGenerator\n\ndatagen = ImageDataGenerator(rotation_range = 10, width_shift_range = 0.1, shear_range = 0.1,\n                             height_shift_range = 0.1, zoom_range = [0.9,1.1], fill_mode = 'nearest', \n                             horizontal_flip = False, vertical_flip = False)\ndatagen.fit(x_train)\n","metadata":{"execution":{"iopub.status.busy":"2026-03-12T14:45:32.521830Z","iopub.execute_input":"2026-03-12T14:45:32.522308Z","iopub.status.idle":"2026-03-12T14:45:32.637110Z","shell.execute_reply.started":"2026-03-12T14:45:32.522261Z","shell.execute_reply":"2026-03-12T14:45:32.636365Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"def Conv_Net_model():\n  model = Sequential()\n  \n  # Block 1\n  model.add(Conv2D(32,(3,3), input_shape=(28,28,1), padding=\"same\", activation=\"relu\"))\n  model.add(BatchNormalization())\n  model.add(Conv2D(32,(3,3), padding=\"same\", activation=\"relu\"))\n  model.add(BatchNormalization())\n  model.add(MaxPooling2D(pool_size=(2,2)))\n  model.add(Dropout(0.25))\n    \n  # Block 2\n  model.add(Conv2D(64,(3,3), activation=\"relu\", padding=\"same\"))\n  model.add(BatchNormalization())\n  model.add(Conv2D(64,(3,3), activation=\"relu\", padding=\"same\"))\n  model.add(BatchNormalization())\n  model.add(MaxPooling2D(pool_size=(2,2)))\n  model.add(Dropout(0.25))\n\n  # Block 3 (Tambahan untuk akurasi tinggi)\n  model.add(Conv2D(128,(3,3), activation=\"relu\", padding=\"same\"))\n  model.add(BatchNormalization())\n  model.add(MaxPooling2D(pool_size=(2,2)))\n  model.add(Dropout(0.25))\n    \n  model.add(Flatten())\n  model.add(Dense(256, activation=\"relu\"))\n  model.add(BatchNormalization())\n  model.add(Dropout(rate=0.4))\n  model.add(Dense(10, activation=\"softmax\"))\n  \n  \n  opt = adam_v2.Adam(0.001, beta_1=0.9, beta_2=0.999, amsgrad=True)\n  model.compile(optimizer=opt, \n                loss=keras.losses.CategoricalCrossentropy(label_smoothing=0.1), \n                metrics=[\"accuracy\"])\n  return model\n\nprint(Conv_Net_model().summary())","metadata":{"execution":{"iopub.status.busy":"2026-03-12T14:45:32.638206Z","iopub.execute_input":"2026-03-12T14:45:32.638497Z","iopub.status.idle":"2026-03-12T14:45:36.251170Z","shell.execute_reply.started":"2026-03-12T14:45:32.638471Z","shell.execute_reply":"2026-03-12T14:45:36.250181Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"nets = 10 \nmodel_ensemble = [0] * nets\n\nreduce_lr = ReduceLROnPlateau(monitor = 'val_accuracy', \n                              min_lr = 0.00001, verbose = 1,  patience = 3, factor = 0.5, mode = 'auto', cooldown = 0)\n\nfor j in range(nets):\n    model_ensemble[j] = Conv_Net_model()\n    print(f\"\\n>>>> TRAINING MODEL KE-{j+1} DARI {nets} <<<<\")\n    model_ensemble[j].fit(datagen.flow(x_train, y_train, batch_size=64), \n                          epochs=45, \n                          validation_data=(x_val, y_val), \n                          shuffle=1, \n                          verbose=1, \n                          callbacks=[reduce_lr])","metadata":{"execution":{"iopub.status.busy":"2026-03-12T14:45:36.252502Z","iopub.execute_input":"2026-03-12T14:45:36.253171Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"y_pred_total = np.zeros((x_test.shape[0], 10))\nfor j in range(nets):\n    y_pred_total += model_ensemble[j].predict(x_test)\n\ny_test = np.argmax(y_pred_total, axis=1)\nprint(y_test.shape)\n","metadata":{"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"sample = pd.read_csv(\"../input/digit-recognizer/sample_submission.csv\")\ni = sample.ImageId\n\noutput = pd.DataFrame({'ImageId' : i,'Label': y_test})\noutput.head(20)","metadata":{"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"display(x_test[18])","metadata":{"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"output.to_csv('submission.csv', index=False)","metadata":{"trusted":true},"outputs":[],"execution_count":null}]}