parm_dict.json

{
  "root_path": "C:/tmp_data/extract_feature/",
  "upload_path": "C:/tmp_data/extract_feature/upload_models/",
  "feature_file_name": "multi_features_data_simu5.pkl",
  "Is_train_autoencoder": false,
  "autoencoder_structure": 4,
  "L1": {
    "noise_threshold": 0.01,
    "_noise_threshold_dsp": "the number less of cluster than the threshold, consider to be noise",
    "choose_model": "Hierarchical",
    "_choose_model_dsp": "options, Hierarchical, DBSCAN",
    "Hierarchical_power_params": [
      [
        0.3,
        3,
        4,
        0.9
      ],
      [
        0.2,
        3,
        5,
        0.9
      ]
    ],
    "_param1_dsp": "param 1 is factor ,the stable indicattor (less than value)(difference from the reverse previous 2 level)",
    "_param2_dsp": "param 2 is min number of cluster",
    "_param3_dsp": "param 3 is max number of cluster",
    "_param4_dsp": "param 4 is standby threshold (percentage)",
    "Is_save_model": true,
    "Hierarchical_vib_params": [
      [
        0.2,
        4,
        10,
        0.5
      ]
    ],
    "_Hierarchical_vib_params_dsp": "For vibration, the format is same",
    "L1_models_name": "L1_models.pkl"
  },
  "L2": {
    "strategy": "consensus",
    "_strategy_dsp": "options: consensus, feature_fusion, meta_clustering, multi_view_cluster",
    "Is_save_model": true,
    "consensus_model": "GMM",
    "_consensus_model_dsp": "model using on each type of sensor clusting: KMeans, GMM",
    "result_type": "prob",
    "_result_type_dsp": "The input the consensus matrix is each submodel result:  prob (probability), one-hot",
    "feature_fusion_pca_factor": 0.95,
    "_feature_fusion_pca_factor_dsp": "how much the PCA catch of the variance\uff0cfor feature_fusion, meta_clustering, multi_view_cluster",
    "feature_fusion_model": "GMM",
    "meta_clustering_model": "GMM",
    "number_class": 0,
    "_number_class_dsp": "The spicify cluster number of each type of sensor. 0 mean search optimal number base on certain score",
    "min_class": 3,
    "_min_class_dsp": "minimum cluster number for search optimal option",
    "max_class": 5,
    "_max_class_dsp": "maximum cluster number (not include) for search optimal option",
    "stable_indicator": 0.3,
    "_stable_indicator_dsp": "when change less than the value, consider it's stable",
    "covariance_type": "diag",
    "_covariance_type_dsp": "GMM parameter, effect on model size"
  }
}