add practice

PA/code/stationary diff.py

@@ -0,0 +1,16 @@
+def difference(timeseries, interval=1):
+    diff = []
+    for i in range(interval, len(timeseries)):
+        value = timeseries[i] - timeseries[i - interval]
+        diff.append(value)
+    return np.array(diff)
+
+diff_ts = difference(trend_ts)
+
+plt.figure(figsize=(12, 6))
+plt.plot(time[1:], diff_ts, label='Äèôôåðåíöèðîâàííûé ðÿä')
+plt.legend()
+plt.title('Ðÿä ïîñëå äèôôåðåíöèðîâàíèÿ')
+plt.show()
+
+adf_test(diff_ts)

PA/code/stationary log.py

@@ -0,0 +1,8 @@
+volatile_ts = np.exp(np.linspace(0, 2, n)) * np.random.normal(size=n)
+
+plt.figure(figsize=(12, 6))
+plt.plot(time, volatile_ts, label='Èñõîäíûé ðÿä')
+plt.plot(time, np.log(volatile_ts), label='Ëîãàðèôìèðîâàííûé ðÿä')
+plt.legend()
+plt.title('Ëîãàðèôìè÷åñêîå ïðåîáðàçîâàíèå')
+plt.show()

PA/code/stationary season.py

@@ -0,0 +1,12 @@
+# Ñîçäàåì ñåçîííûé ðÿä
+seasonal_ts = 5 * np.sin(np.linspace(0, 10*np.pi, n)) + np.linspace(0, 5, n) + np.random.normal(size=n)
+
+# Ñåçîííîå äèôôåðåíöèðîâàíèå ñ ïåðèîäîì 30 äíåé
+seasonal_diff = seasonal_ts[30:] - seasonal_ts[:-30]
+
+plt.figure(figsize=(12, 6))
+plt.plot(time, seasonal_ts, label='Èñõîäíûé ðÿä')
+plt.plot(time[30:], seasonal_diff, label='Ñåçîííî äèôôåðåíöèðîâàííûé ðÿä')
+plt.legend()
+plt.title('Ñåçîííîå äèôôåðåíöèðîâàíèå')
+plt.show()

PA/code/stationary test.py

@@ -0,0 +1,39 @@
+import pandas as pd
+import matplotlib.pyplot as plt
+import numpy as np
+from statsmodels.tsa.stattools import adfuller, kpss
+
+
+# Ïðèìåð âðåìåííîãî ðÿäà
+np.random.seed(42)
+n = 100
+time = pd.date_range(start='2020-01-01', periods=n, freq='D')
+stationary_ts = np.random.normal(loc=0, scale=1, size=n)
+trend_ts = np.linspace(0, 10, n) + stationary_ts
+
+plt.figure(figsize=(12, 6))
+plt.plot(time, stationary_ts, label='Ñòàöèîíàðíûé ðÿä')
+plt.plot(time, trend_ts, label='Íåñòàöèîíàðíûé ðÿä ñ òðåíäîì')
+plt.legend()
+plt.title('Âèçóàëüíàÿ ïðîâåðêà ñòàöèîíàðíîñòè')
+plt.show()
+
+def adf_test(timeseries):
+    print("Ðåçóëüòàòû òåñòà Äèêè-Ôóëëåðà:")
+    dftest = adfuller(timeseries, autolag='AIC')
+    dfoutput = pd.Series(dftest[0:4],
+                         index=['Òåñòîâàÿ ñòàòèñòèêà', 'p-value',
+                                'Êîëè÷åñòâî ëàãîâ', 'Êîëè÷åñòâî íàáëþäåíèé'])
+    for key, value in dftest[4].items():
+        dfoutput[f'Êðèòè÷åñêîå çíà÷åíèå ({key})'] = value
+    print(dfoutput)
+
+    if dftest[1] <= 0.05:
+        print("Âûâîä: ðÿä ñòàöèîíàðíûé (îòâåðãàåì H0)")
+    else:
+        print("Âûâîä: ðÿä íåñòàöèîíàðíûé (íå îòâåðãàåì H0)")
+
+
+adf_test(stationary_ts)
+print("\n")
+adf_test(trend_ts)