Python/machine_learning/forecasting/run.py

158 lines
5.5 KiB
Python
Raw Normal View History

"""
this is code for forecasting
but i modified it and used it for safety checker of data
for ex: you have a online shop and for some reason some data are
missing (the amount of data that u expected are not supposed to be)
then we can use it
*ps : 1. ofc we can use normal statistic method but in this case
the data is quite absurd and only a little^^
2. ofc u can use this and modified it for forecasting purpose
for the next 3 months sales or something,
u can just adjust it for ur own purpose
"""
import numpy as np
import pandas as pd
from sklearn.preprocessing import Normalizer
from sklearn.svm import SVR
from statsmodels.tsa.statespace.sarimax import SARIMAX
def linear_regression_prediction(
train_dt: list, train_usr: list, train_mtch: list, test_dt: list, test_mtch: list
) -> float:
"""
First method: linear regression
input : training data (date, total_user, total_event) in list of float
output : list of total user prediction in float
>>> n = linear_regression_prediction([2,3,4,5], [5,3,4,6], [3,1,2,4], [2,1], [2,2])
>>> abs(n - 5.0) < 1e-6 # Checking precision because of floating point errors
True
"""
x = [[1, item, train_mtch[i]] for i, item in enumerate(train_dt)]
x = np.array(x)
y = np.array(train_usr)
beta = np.dot(np.dot(np.linalg.inv(np.dot(x.transpose(), x)), x.transpose()), y)
return abs(beta[0] + test_dt[0] * beta[1] + test_mtch[0] + beta[2])
def sarimax_predictor(train_user: list, train_match: list, test_match: list) -> float:
"""
second method: Sarimax
sarimax is a statistic method which using previous input
and learn its pattern to predict future data
input : training data (total_user, with exog data = total_event) in list of float
output : list of total user prediction in float
>>> sarimax_predictor([4,2,6,8], [3,1,2,4], [2])
6.6666671111109626
"""
order = (1, 2, 1)
seasonal_order = (1, 1, 0, 7)
model = SARIMAX(
train_user, exog=train_match, order=order, seasonal_order=seasonal_order
)
model_fit = model.fit(disp=False, maxiter=600, method="nm")
result = model_fit.predict(1, len(test_match), exog=[test_match])
return result[0]
def support_vector_regressor(x_train: list, x_test: list, train_user: list) -> float:
"""
Third method: Support vector regressor
svr is quite the same with svm(support vector machine)
it uses the same principles as the SVM for classification,
with only a few minor differences and the only different is that
it suits better for regression purpose
input : training data (date, total_user, total_event) in list of float
where x = list of set (date and total event)
output : list of total user prediction in float
>>> support_vector_regressor([[5,2],[1,5],[6,2]], [[3,2]], [2,1,4])
1.634932078116079
"""
regressor = SVR(kernel="rbf", C=1, gamma=0.1, epsilon=0.1)
regressor.fit(x_train, train_user)
y_pred = regressor.predict(x_test)
return y_pred[0]
def interquartile_range_checker(train_user: list) -> float:
"""
Optional method: interquatile range
input : list of total user in float
output : low limit of input in float
this method can be used to check whether some data is outlier or not
>>> interquartile_range_checker([1,2,3,4,5,6,7,8,9,10])
2.8
"""
train_user.sort()
q1 = np.percentile(train_user, 25)
q3 = np.percentile(train_user, 75)
iqr = q3 - q1
low_lim = q1 - (iqr * 0.1)
return low_lim
def data_safety_checker(list_vote: list, actual_result: float) -> None:
"""
Used to review all the votes (list result prediction)
and compare it to the actual result.
input : list of predictions
output : print whether it's safe or not
>>> data_safety_checker([2,3,4],5.0)
Today's data is not safe.
"""
safe = 0
not_safe = 0
for i in list_vote:
if i > actual_result:
safe = not_safe + 1
else:
if abs(abs(i) - abs(actual_result)) <= 0.1:
safe = safe + 1
else:
not_safe = not_safe + 1
print(f"Today's data is {'not ' if safe <= not_safe else ''}safe.")
# data_input_df = pd.read_csv("ex_data.csv", header=None)
data_input = [[18231, 0.0, 1], [22621, 1.0, 2], [15675, 0.0, 3], [23583, 1.0, 4]]
data_input_df = pd.DataFrame(data_input, columns=["total_user", "total_even", "days"])
"""
data column = total user in a day, how much online event held in one day,
what day is that(sunday-saturday)
"""
# start normalization
normalize_df = Normalizer().fit_transform(data_input_df.values)
# split data
total_date = normalize_df[:, 2].tolist()
total_user = normalize_df[:, 0].tolist()
total_match = normalize_df[:, 1].tolist()
# for svr (input variable = total date and total match)
x = normalize_df[:, [1, 2]].tolist()
x_train = x[: len(x) - 1]
x_test = x[len(x) - 1 :]
# for linear reression & sarimax
trn_date = total_date[: len(total_date) - 1]
trn_user = total_user[: len(total_user) - 1]
trn_match = total_match[: len(total_match) - 1]
tst_date = total_date[len(total_date) - 1 :]
tst_user = total_user[len(total_user) - 1 :]
tst_match = total_match[len(total_match) - 1 :]
# voting system with forecasting
res_vote = []
res_vote.append(
linear_regression_prediction(trn_date, trn_user, trn_match, tst_date, tst_match)
)
res_vote.append(sarimax_predictor(trn_user, trn_match, tst_match))
res_vote.append(support_vector_regressor(x_train, x_test, trn_user))
# check the safety of todays'data^^
data_safety_checker(res_vote, tst_user)