es = keras.callbacks.EarlyStopping(patience=n_stop, restore_best_weights=True)\nlr = keras.callbacks.LearningRateScheduler(scheduler)\n\ncv = KFold(n_splits=n_fold, shuffle=True, random_state=seed)\n\nP = np.zeros_like(Y, dtype=float)\nP_tst = np.zeros((X_tst.shape[0], output_dim), dtype=float)\n\nfor i, (i_trn, i_val) in enumerate(cv.split(X), 1):\n print(f'Training for CV #{i}')\n model = build_model()\n history = model.fit(X[i_trn], Y[i_trn], validation_data=(X[i_val], Y[i_val]), \n epochs=epochs, batch_size=batch_size, callbacks=[es, lr], verbose=0)\n P[i_val] = label_scaler.inverse_transform(model.predict(X[i_val]))\n P_tst += label_scaler.inverse_transform(model.predict(X_tst)) / n_fold\n \n distance_i = calc_haversine(df.latDeg_gt.values[i_val], \n df.lngDeg_gt.values[i_val], \n P[i_val, 0] + df.latDeg.values[i_val], \n P[i_val, 1] + df.lngDeg.values[i_val]).mean()\n print(f'CV #{i}: {np.percentile(distance_i, [50, 95])}')\n","metadata":{"_kg_hide-output":true,"execution":{"iopub.status.busy":"2021-07-08T11:53:03.029571Z","iopub.execute_input":"2021-07-08T11:53:03.030109Z","iopub.status.idle":"2021-07-08T11:58:45.648201Z","shell.execute_reply.started":"2021-07-08T11:53:03.030066Z","shell.execute_reply":"2021-07-08T11:58:45.647094Z"}}},{"cell_type":"code","source":"print(P.mean(axis=0), P_tst.mean(axis=0))\nnp.savetxt(predict_val_file, P, delimiter=',', fmt='%.6f')\nnp.savetxt(predict_tst_file, P_tst, delimiter=',', fmt='%.6f')","metadata":{"execution":{"iopub.status.busy":"2021-07-24T00:32:47.045487Z","iopub.execute_input":"2021-07-24T00:32:47.046033Z","iopub.status.idle":"2021-07-24T00:32:47.808976Z","shell.execute_reply.started":"2021-07-24T00:32:47.045988Z","shell.execute_reply":"2021-07-24T00:32:47.808109Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"distance = calc_haversine(df.latDeg_gt, df.lngDeg_gt, P[:, 0] + df.latDeg, P[:, 1] + df.lngDeg)\nprint(f'CV All: {np.percentile(distance, [50, 95])}')","metadata":{"execution":{"iopub.status.busy":"2021-07-24T00:32:47.809926Z","iopub.execute_input":"2021-07-24T00:32:47.810203Z","iopub.status.idle":"2021-07-24T00:32:47.835409Z","shell.execute_reply.started":"2021-07-24T00:32:47.810177Z","shell.execute_reply":"2021-07-24T00:32:47.834537Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"df.sort_values([phone_col, ts_col], inplace=True)\ndf_smoothed = df.copy()\ndf_smoothed[lat_col] = df[lat_col] + P[:, 0]\ndf_smoothed[lon_col] = df[lon_col] + P[:, 1]\ndf_smoothed = apply_kf_smoothing(df_smoothed)\ndistance = calc_haversine(df_smoothed.latDeg_gt, df_smoothed.lngDeg_gt, df_smoothed.latDeg, df_smoothed.lngDeg)\nprint(f'CV All (smoothed): {np.percentile(distance, [50, 95])}')","metadata":{"execution":{"iopub.status.busy":"2021-07-24T00:32:47.836593Z","iopub.execute_input":"2021-07-24T00:32:47.836872Z","iopub.status.idle":"2021-07-24T00:33:39.896561Z","shell.execute_reply.started":"2021-07-24T00:32:47.836845Z","shell.execute_reply":"2021-07-24T00:33:39.895642Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# Submission File","metadata":{}},{"cell_type":"code","source":"distance_tst = calc_haversine(df_tst.latDeg, df_tst.lngDeg, P_tst[:, 0] + df_tst.latDeg, P_tst[:, 1] + df_tst.lngDeg)\nprint(f'CV All: {np.percentile(distance_tst, [50, 95])}')","metadata":{"execution":{"iopub.status.busy":"2021-07-24T00:33:39.898024Z","iopub.execute_input":"2021-07-24T00:33:39.898422Z","iopub.status.idle":"2021-07-24T00:33:39.922169Z","shell.execute_reply.started":"2021-07-24T00:33:39.898375Z","shell.execute_reply":"2021-07-24T00:33:39.921373Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"numpy.percentile()は、与えた配列から指定したパーセンタイル値を計算する。\n\npercentile(a, q)\na：パーセンタイルを計算する元の配列。\nq：パーセンタイル値、または配列。パーセンタイル値は0～100で、百分率表示であることに注意。1次元配列を指定すると、各要素のパーセンタイル値に相当する値が同じサイズの配列で返される。\n\nprint(np.percentile(a, [55, 75]))\n [5.5 7.5]","metadata":{}},{"cell_type":"code","source":"distance_tst","metadata":{"execution":{"iopub.status.busy":"2021-07-24T00:33:39.923461Z","iopub.execute_input":"2021-07-24T00:33:39.923767Z","iopub.status.idle":"2021-07-24T00:33:39.931518Z","shell.execute_reply.started":"2021-07-24T00:33:39.923739Z","shell.execute_reply":"2021-07-24T00:33:39.930446Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"df_tst.sort_values([phone_col, ts_col], inplace=True)\ndf_tst_smoothed = df_tst.copy()\ndf_tst_smoothed[lat_col] = df_tst_smoothed[lat_col] + P_tst[:, 0]\ndf_tst_smoothed[lon_col] = df_tst_smoothed[lon_col] + P_tst[:, 1]\ndf_tst_smoothed","metadata":{"execution":{"iopub.status.busy":"2021-07-24T00:33:39.933157Z","iopub.execute_input":"2021-07-24T00:33:39.93374Z","iopub.status.idle":"2021-07-24T00:33:40.062347Z","shell.execute_reply.started":"2021-07-24T00:33:39.933695Z","shell.execute_reply":"2021-07-24T00:33:40.061638Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"df_tst_smoothed = apply_kf_smoothing(df_tst_smoothed)\ndf_tst_smoothed","metadata":{"execution":{"iopub.status.busy":"2021-07-24T00:33:40.063299Z","iopub.execute_input":"2021-07-24T00:33:40.063686Z","iopub.status.idle":"2021-07-24T00:34:14.356418Z","shell.execute_reply.started":"2021-07-24T00:33:40.063659Z","shell.execute_reply":"2021-07-24T00:34:14.355717Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"distance_tst = calc_haversine(df_tst.latDeg, df_tst.lngDeg, df_tst_smoothed.latDeg, df_tst_smoothed.lngDeg)\nprint(f'CV All (smoothed): {np.percentile(distance_tst, [50, 95])}')","metadata":{"execution":{"iopub.status.busy":"2021-07-24T00:34:14.35747Z","iopub.execute_input":"2021-07-24T00:34:14.357915Z","iopub.status.idle":"2021-07-24T00:34:14.377429Z","shell.execute_reply.started":"2021-07-24T00:34:14.357863Z","shell.execute_reply":"2021-07-24T00:34:14.376734Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"df_tst_smoothed[[phone_col, ts_col, lat_col, lon_col]].to_csv(submission_file, index=False)","metadata":{"execution":{"iopub.status.busy":"2021-07-24T00:34:14.378557Z","iopub.execute_input":"2021-07-24T00:34:14.379003Z","iopub.status.idle":"2021-07-24T00:34:15.05731Z","shell.execute_reply.started":"2021-07-24T00:34:14.378962Z","shell.execute_reply":"2021-07-24T00:34:15.056314Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"submission = df_tst_smoothed[[phone_col, ts_col, lat_col, lon_col]]\n","metadata":{"execution":{"iopub.status.busy":"2021-07-24T00:34:15.0587Z","iopub.execute_input":"2021-07-24T00:34:15.059024Z","iopub.status.idle":"2021-07-24T00:34:15.066667Z","shell.execute_reply.started":"2021-07-24T00:34:15.058995Z","shell.execute_reply":"2021-07-24T00:34:15.065614Z"},"trusted":true},"execution_count":null,"outputs":[]}]}