103 lines
3.7 KiB
Python
103 lines
3.7 KiB
Python
import numpy as np
|
|
from sklearn.preprocessing import StandardScaler
|
|
from sklearn.metrics import mean_squared_error
|
|
from sklearn.model_selection import train_test_split
|
|
import v2realbot.ml.mlutils as mu
|
|
from keras.layers import LSTM, Dense
|
|
import matplotlib.pyplot as plt
|
|
from v2realbot.ml.ml import ModelML
|
|
from v2realbot.enums.enums import PredOutput, Source, TargetTRFM
|
|
from v2realbot.controller.services import get_archived_runner_details_byID, update_archive_detail
|
|
# from collections import defaultdict
|
|
# from operator import itemgetter
|
|
from joblib import load
|
|
|
|
#TODO - DO API
|
|
# v ml atomicke api pro evaluaci (runneru, batche)
|
|
# v services: model.add_vector_prediction_to_archrunner_as_new_indicator (vrátí v podstate obohacený archDetail) - nebo i ukládat do db? uvidime
|
|
# v rest api prevolani
|
|
# db support: zatim jen ciselnik modelu + jeho zakladni nastaveni, obrabeci api, load modelu zatim z file
|
|
|
|
cfg: ModelML = mu.load_model("model1", "0.1")
|
|
|
|
|
|
#EVALUATE SPECIFIC RUNNER - VECTOR BASED (toto dat do samostatne API pripadne pak udelat nadstavnu na batch a runners)
|
|
#otestuje model na neznamem runnerovi, seznamu runneru nebo batch_id
|
|
|
|
|
|
|
|
runner_id = "a38fc269-8df3-4374-9506-f0280d798854"
|
|
save_new_ind = True
|
|
source_data, target_data, rows_in_day = cfg.load_data(runners_ids=[runner_id])
|
|
|
|
if len(rows_in_day) > 1:
|
|
#pro vis se cela tato sluzba volat v loopu
|
|
raise Exception("Vytvareni indikatoru dostupne zatim jen pro jeden runner")
|
|
|
|
#scalujeme X
|
|
source_data = cfg.scalerX.fit_transform(source_data)
|
|
|
|
#tady si vyzkousim i skrz vice runneru
|
|
X_eval, y_eval, y_eval_ref = cfg.create_sequences(combined_data=source_data, target_data=target_data,remove_cross_sequences=True, rows_in_day=rows_in_day)
|
|
|
|
#toto nutne?
|
|
X_eval = np.array(X_eval)
|
|
y_eval = np.array(y_eval)
|
|
y_eval_ref = np.array(y_eval_ref)
|
|
#scaluji target - nemusis
|
|
#y_eval = cfg.scalerY.fit_transform(y_eval)
|
|
|
|
X_eval = cfg.model.predict(X_eval)
|
|
X_eval = cfg.scalerY.inverse_transform(X_eval)
|
|
print("po predikci x_eval shape", X_eval.shape)
|
|
|
|
#pokud mame dostupnou i target v runneru, pak pridame porovnavaci indikator
|
|
difference_mse = None
|
|
if len(y_eval) > 0:
|
|
#TODO porad to pliva 1 hodnotu
|
|
difference_mse = mean_squared_error(y_eval, X_eval,multioutput="raw_values")
|
|
|
|
print("ted mam tedy dva nove sloupce")
|
|
print("X_eval", X_eval.shape)
|
|
if difference_mse is not None:
|
|
print("difference_mse", difference_mse.shape)
|
|
print(f"zplostime je, dopredu pridame {cfg.input_sequences-1} a dozadu nic")
|
|
#print(f"a melo by nam to celkem dat {len(bars['time'])}")
|
|
#tohle pak nejak doladit, ale vypada to good
|
|
#plus do druheho indikatoru pridat ten difference_mse
|
|
|
|
#TODO jeste je posledni hodnota predikce nejak OFF (2.52... ) - podivat se na to
|
|
#TODO na produkci srovnat se skutecnym BT predictem (mozna zde bude treba seq-1) -
|
|
# prvni predikce nejspis uz bude na desítce
|
|
ind_pred = list(np.concatenate([np.zeros(cfg.input_sequences-1), X_eval.ravel()]))
|
|
print(ind_pred)
|
|
print(len(ind_pred))
|
|
print("tada")
|
|
#ted k nim pridame
|
|
|
|
if save_new_ind:
|
|
#novy ind ulozime do archrunnera (na produkci nejspis jen show)
|
|
res, sada = get_archived_runner_details_byID(runner_id)
|
|
if res == 0:
|
|
print("ok")
|
|
else:
|
|
print("error",res,sada)
|
|
raise Exception(f"error loading runner {runner_id} : {res} {sada}")
|
|
|
|
sada["indicators"][0]["pred_added"] = ind_pred
|
|
|
|
req, res = update_archive_detail(runner_id, sada)
|
|
print(f"indicator pred_added was ADDED to {runner_id}")
|
|
|
|
|
|
# Plot the predicted vs. actual
|
|
plt.plot(y_eval, label='Target')
|
|
plt.plot(X_eval, label='Predicted')
|
|
#TODO zde nejak vymyslet jinou pricelinu - jako lightweight chart
|
|
if difference_mse is not None:
|
|
plt.plot(difference_mse, label='diference')
|
|
plt.plot(y_eval_ref, label='reference column - vwap')
|
|
plt.plot()
|
|
plt.legend()
|
|
plt.show()
|