Update app.py

app.py

@@ -1,32 +1,18 @@
 import gradio as gr
 import pandas as pd
 import numpy as np
-from statsmodels.tsa.seasonal import STL
-from pmdarima import auto_arima
 import matplotlib.pyplot as plt
 
-def detect_periodicity(series):
-    # Use o método auto_arima para determinar a periodicidade da série
-    stepwise_model = auto_arima(series, start_p=1, start_q=1, max_p=3, max_q=3, m=12, seasonal=True, trace=True)
-    return stepwise_model.seasonal_order[0]
+def detect_anomalies(series, window_size=7, threshold=2.5):
+    # Calculate the rolling mean and standard deviation using the specified window size
+    rolling_mean = series.rolling(window=window_size, center=True).mean()
+    rolling_std = series.rolling(window=window_size, center=True).std()
 
-def detect_anomalies(series, seasonality_window=7, anomaly_threshold=0.05):
-    # Decompose series using Seasonal-Trend decomposition using LOESS (STL)
-    decomposition = STL(series, seasonal=seasonality_window).fit()
-    trend = decomposition.trend
-    seasonal = decomposition.seasonal
-    residual = decomposition.resid
+    # Calculate the z-score for each data point
+    z_scores = np.abs((series - rolling_mean) / rolling_std)
 
-    # Combine trend and seasonal components to form detrended series
-    detrended_series = series - trend - seasonal
-
-    # Fit the Isolation Forest model to detect anomalies in the detrended series
-    model = IsolationForest(contamination=anomaly_threshold)
-    model.fit(detrended_series.values.reshape(-1, 1))
-
-    # Predict outliers in the detrended series
-    outlier_predictions = model.predict(detrended_series.values.reshape(-1, 1))
-    anomalies = detrended_series[outlier_predictions == -1]
+    # Detect anomalies based on the threshold
+    anomalies = series[z_scores > threshold]
 
     return anomalies.index
 
@@ -41,17 +27,14 @@ def plot_anomalies(series, anomalies):
     plt.show()
 
 # Função para o Gradio
-def anomaly_detection_app(time_series_txt):
-    # Carregar a série temporal a partir do arquivo de texto
-    df = pd.read_csv(time_series_txt, header=None, names=["Date", "Value"])
+def anomaly_detection_app(time_series_csv, window_size=7, threshold=2.5):
+    # Carregar a série temporal a partir do arquivo CSV
+    df = pd.read_csv(time_series_csv)
     df['Date'] = pd.to_datetime(df['Date'])
     df = df.set_index('Date')
 
-    # Determinar a periodicidade da série automaticamente
-    periodicity = detect_periodicity(df['Value'])
-
-    # Detectar anomalias na série temporal
-    anomalies = detect_anomalies(df['Value'], seasonality_window=periodicity)
+    # Detectar anomalias na série temporal usando a média móvel
+    anomalies = detect_anomalies(df['Value'], window_size=window_size, threshold=threshold)
 
     # Plotar a série temporal com as anomalias destacadas
     plot_anomalies(df['Value'], anomalies)
@@ -61,10 +44,14 @@ def anomaly_detection_app(time_series_txt):
 # Interface do Gradio
 iface = gr.Interface(
     fn=anomaly_detection_app,
-    inputs=gr.inputs.File(label="Carregar série temporal (TXT)"),
+    inputs=[
+        gr.inputs.File(label="Carregar série temporal (CSV)"),
+        gr.inputs.Number(default=7, label="Tamanho da janela da média móvel"),
+        gr.inputs.Number(default=2.5, label="Threshold para anomalias (em desvios-padrão)")
+    ],
     outputs="text",
     title="Análise de Anomalias em Séries Temporais",
-    description="Este aplicativo detecta anomalias em uma série temporal carregada através de um arquivo de texto (TXT).",
+    description="Este aplicativo detecta anomalias em uma série temporal carregada através de um arquivo CSV usando a média móvel.",
     live=True
 )