How to Use?

This is a project for the 2023 study of the Sensorium research project. \ All copyrights belong to Sanghoon Lee, Ji-yeong Chae. \ For questions, please email leesh2913@dgist.ac.kr.

preprocess.py

process raw file to int csv file
csv file is saved at data/x_data_ .csv and data/y_data_ .csv

pre_process_00

save only frequency (0 ~ 5120hz data)
data name is data/x_data_00.csv, data/y_data_00.csv

pre_process_01

frequency (0 ~ 5120hz data) + Pipe info (p pd s type data) data name is data/x_data_01.csv, data/y_data_01.csv

train.py

1. import train.py

import train

2. Initialize train class

trainer = train.train(name)

name is the file name of /data folder example: "00", "01"

3. search_parameter(model_name, rand_seed, parameter, name)

params = {'n_neighbors': [5, 15, 30, 100], 'weights':  ['uniform', 'distance'],
          'algorithm': ['auto', 'ball_tree', 'kd_tree', 'brute']}

trainer.search_parameter("KNN", 3, param, "First test")

model_name is the name of ML methods
examples: "KNN", "MLP", "DNN", "GB", etc ...
rand_seed is the seed of randome state examples: 0, 42, 1040, etc ...

parameter is the dictionary of parameter set examples:

params = {
    'hidden_layer_sizes': [(50,), (100,), (150,)],
    'activation': ['tanh', 'relu', 'logistic'],
    'solver': ['adam'],
    'alpha': [0.001, 0.01, 0.1]
}

You can check the parameter settings in parameter_list.txt

name is the file name that you want to save examples: 0234
If model_name is MLP and name is "All" then
the result of GridSearchCV is saved at log/MLP_ALL.csv

Parameter_list.txt

Save the possible parameter list of any M.L methods.

cobaal/sensorium