def build_ar_models_for_traces(trace_lst, results_dict, train_prop,
val_prop, max_mem):
fixed_model_params = {'train_prop': train_prop,
'val_prop': val_prop,
'max_mem': max_mem}
ar_params_lst = [{'p': p, **fixed_model_params} for p in AR_COMPS]
analysis.get_best_model_results_for_traces(
TraceAR, ar_params_lst, trace_lst,
results_dict, specs.MODELS_COUNT, fixed_model_params)
def build_es_models_for_traces(trace_lst, results_dict,
train_prop, val_prop, max_mem):
fixed_model_params = {'initial_pred': 0.001, 'train_prop': train_prop,
'val_prop': val_prop, 'max_mem': max_mem}
es_params_lst = [{'alpha': alpha_val, **fixed_model_params}
for alpha_val in ALPHAS]
analysis.get_best_model_results_for_traces(
TraceExponentialSmoothing, es_params_lst, trace_lst,
results_dict, specs.MODELS_COUNT, fixed_model_params)
def build_ma_models_for_traces(trace_lst, results_dict, train_prop,
val_prop, max_mem):
fixed_model_params = {'initial_pred': 0.0, 'train_prop': train_prop,
'val_prop': val_prop, 'max_mem': max_mem}
ma_params_lst = [{'window_length': ma_win, **fixed_model_params}
for ma_win in MA_WINDOWS]
analysis.get_best_model_results_for_traces(
TraceMovingAverage, ma_params_lst, trace_lst,
results_dict, specs.MODELS_COUNT, fixed_model_params)
def build_arima_models_for_traces(traces_lst, results_dict,
train_prop, val_prop, max_mem):
fixed_model_params = {'train_prop': train_prop,
'val_prop': val_prop,
'max_mem': max_mem}
arima_params_lst = [{'p': p, 'd': d, 'q': q, **fixed_model_params}
for p, d, q in ARIMA_PARAMS]
analysis.get_best_model_results_for_traces(
TraceARIMA, arima_params_lst, traces_lst,
results_dict, specs.MODELS_COUNT, fixed_model_params)
def build_arimax_models_for_traces(traces_lst, results_dict,
train_prop, val_prop, max_mem):
target_col = specs.get_target_variable(max_mem)
data_handler = MLDataHandler(
FEATURE_COLS, [target_col], train_prop, val_prop)
arimax_params_lst = [{'data_handler': data_handler, 'lags': specs.LAGS,
'p': p, 'd': d, 'q': q}
for p, d, q in product(ARIMA_p, ARIMA_d, ARIMA_q)]
analysis.get_best_model_results_for_traces(
TraceARIMAX, arimax_params_lst, traces_lst,
results_dict, specs.MODELS_COUNT)
def build_reg_models_for_traces(trace_lst, results_dict,
train_prop, val_prop, max_mem):
target_col = specs.get_target_variable(max_mem)
data_handler = MLDataHandler(
FEATURE_COLS, [target_col], train_prop, val_prop)
fixed_model_params = {'data_handler': data_handler, 'lags': specs.LAGS}
reg_params_lst = [{'reg_val': reg_val, **fixed_model_params}
for reg_val in REG_VALS]
analysis.get_best_model_results_for_traces(
TraceRegression, reg_params_lst, trace_lst,
results_dict, specs.MODELS_COUNT, fixed_model_params)
def build_xgb_models_for_traces(trace_lst, results_dict,
train_prop, val_prop, max_mem):
target_col = specs.get_target_variable(max_mem)
data_handler = MLDataHandler(
FEATURE_COLS, [target_col], train_prop, val_prop)
fixed_model_params = {'data_handler': data_handler, 'lags': specs.LAGS}
xgb_params_lst = [{'learning_rate': learning_rate,
'estimators': n_estimators,
'depth': depth,
**fixed_model_params}
for learning_rate, n_estimators, depth
in product(LEARNING_RATES, ESTIMATORS, DEPTHS)]
analysis.get_best_model_results_for_traces(
TraceXGB, xgb_params_lst, trace_lst,
results_dict, specs.MODELS_COUNT, fixed_model_params)
