Update app.py

app.py

@@ -3,8 +3,7 @@ import numpy as np
 import pandas as pd
 import plotly.express as px
 import plotly.graph_objs as go
-from statsforecast import StatsForecast
-from statsforecast.models import auto_arima
+from pmdarima import auto_arima
 from utils import calc_seasonality
 
 @st.cache
@@ -19,9 +18,9 @@ if uploaded_file is not None:
      st.write(dataframe)
      series = st.text_input("Write the name of the variable you want to forecast")
      date = st.text_input('Write the first date')
-#     freq = st.text_input('Write the frequency')
+     # freq = st.text_input('Write the frequency')
      freq = st.selectbox('Select the frequency', 
-                          ('Y', 'Q', 'M', 'W', 'D'))
+                          ('Y', 'Q', 'M', 'W', 'D', 'NULL'))
               
      if series is not None and freq is not None:
 
@@ -31,27 +30,43 @@ if uploaded_file is not None:
 
      
           series_train = pd.DataFrame(
-                                        {
-                                         'ds': pd.date_range(start=date, periods=dataframe.shape[0], freq=freq),
-                                         'y': dataframe[series].values
-                                        },
-                                        index=pd.Index([0] * dataframe.shape[0], name='unique_id')
-                                      )
+                                            {
+                                             'ds': pd.date_range(start=date, periods=dataframe.shape[0], freq=freq),
+                                             'y': dataframe[series].values
+                                            },
+                                            index=pd.Index([0] * dataframe.shape[0], name='unique_id')
+                                          )              
               
           fig = px.line(series_train, x='ds', y='y', title=f'{series}')
           st.plotly_chart(fig)
           with st.spinner('Wait for it...'):
-              fcst = StatsForecast(
-                                   series_train, 
-                                   models=[(auto_arima, seasonality), ], 
-                                   freq=freq, 
-                                   n_jobs=1
-                                  )
-              forecasts = fcst.forecast(horizons, level=(80, 95))  
+              if freq != 'NULL':
+
+                  fcst = auto_arima(
+                                       series_train, 
+                                       seasonality=True, 
+                                       freq=freq, 
+                                       n_jobs=1,
+                                       max_p=12,
+                                       max_q=12
+                                      )
+              else:
+                  fcst = auto_arima(
+                                       series_train, 
+                                       seasonality=False, 
+                                       n_jobs=1,
+                                       max_p=12,
+                                       max_q=12
+                                      )
+              fcst.fit(series_train)    
+              forecasts, ci_05 = fcst.predict(horizons, alpha=0.05)
+              _, ci_10 = fcst.predict(horizons, alpha=0.1)
           st.success('Done!')
             
           st.write(forecasts)
-          
+                  forecasts_df = pd.DataFrame({'ds':pd.date_range(start=serie['ds']+pd.to_timedelta(horizons, units=freq), periods=horizons, freq=freq),
+                                       'mean':forecasts, 'low_ci_05':ci_05[:,0], 'low_ci_10':ci_10[:,0],
+                                       'hi_ci_05':ci_05[:,1], 'hi_ci_10':ci_10[:,1]})
           csv = convert_df(forecasts)
 
           st.download_button(
@@ -74,7 +89,7 @@ if uploaded_file is not None:
                        go.Scatter(
                            name=f'auto_arima_season_length-{seasonality}_mean',
                            x=forecasts['ds'],
-                           y=forecasts[f'auto_arima_season_length-{seasonality}_mean'],
+                           y=forecasts['mean'],
                            mode='lines',
                            marker=dict(color='green'),
                            line=dict(width=1),