Files
Time-Series-Analysis/models/plotting.py
2025-08-01 04:33:03 -04:00

138 lines
5.1 KiB
Python
Executable File

import pandas as pd
import plotly.express as px
import plotly.graph_objects as go
import plotly.io as pio
from statsmodels.graphics.tsaplots import plot_acf, plot_pacf
import matplotlib
matplotlib.use('Agg')
import matplotlib.pyplot as plt
import io
import base64
from statsmodels.tsa.holtwinters import ExponentialSmoothing
import pmdarima as pm
from prophet import Prophet
def create_acf_pacf_plots(data):
# Create ACF and PACF plots using matplotlib
fig, (ax1, ax2) = plt.subplots(2, 1, figsize=(10, 8))
plot_acf(data, ax=ax1, lags=40)
ax1.set_title('Autocorrelation Function')
plot_pacf(data, ax=ax2, lags=40)
ax2.set_title('Partial Autocorrelation Function')
# Convert matplotlib plot to Plotly
buf = io.BytesIO()
plt.savefig(buf, format='png')
plt.close(fig)
buf.seek(0)
img_str = base64.b64encode(buf.getvalue()).decode('utf-8')
# Create Plotly figure with image
fig_plotly = go.Figure()
fig_plotly.add_layout_image(
dict(
source=f'data:image/png;base64,{img_str}',
x=0,
y=1,
xref="paper",
yref="paper",
sizex=1,
sizey=1,
sizing="stretch",
opacity=1,
layer="below"
)
)
fig_plotly.update_layout(
height=600,
showlegend=False,
xaxis=dict(visible=False),
yaxis=dict(visible=False)
)
return pio.to_html(fig_plotly, full_html=False)
def create_comparison_plot(filepath, forecast_history, selected_indices):
# Read data
if filepath.endswith('.csv'):
df = pd.read_csv(filepath)
else:
df = pd.read_excel(filepath)
date_col = df.columns[0]
value_col = df.columns[1]
df[date_col] = pd.to_datetime(df[date_col])
df.set_index(date_col, inplace=True)
# Create Plotly figure
fig = go.Figure()
fig.add_trace(go.Scatter(x=df.index, y=df[value_col], name='Historical', line=dict(color='black')))
# Use Plotly qualitative colors
colors = px.colors.qualitative.Plotly
# Generate forecasts for selected indices
for idx, run_idx in enumerate(selected_indices):
entry = forecast_history[run_idx]
train_percent = entry['train_percent'] / 100
forecast_periods = entry['forecast_periods']
model_type = entry['model_type']
# Split data
train_size = int(len(df) * train_percent)
test_size = len(df) - train_size
train_data = df[value_col].iloc[:train_size]
test_data = df[value_col].iloc[train_size:] if test_size > 0 else pd.Series()
forecast_dates = pd.date_range(start=df.index[-1], periods=forecast_periods + 1, freq=df.index.inferred_freq)[
1:]
# Run model based on model_type
forecast = None
if model_type == 'ARIMA':
model = pm.auto_arima(train_data,
seasonal=True,
m=12,
start_p=0, start_q=0,
max_p=3, max_q=3,
start_P=0, start_Q=0,
max_P=2, max_Q=2,
d=1, D=1,
trace=False,
error_action='ignore',
suppress_warnings=True,
stepwise=True)
model_fit = model.fit(train_data)
forecast = model_fit.predict(n_periods=forecast_periods)
elif model_type == 'Exponential Smoothing':
model = ExponentialSmoothing(train_data,
trend='add',
seasonal='add',
seasonal_periods=12)
model_fit = model.fit()
forecast = model_fit.forecast(forecast_periods)
elif model_type == 'Prophet':
prophet_df = train_data.reset_index().rename(columns={date_col: 'ds', value_col: 'y'})
model = Prophet(yearly_seasonality=True, weekly_seasonality=False, daily_seasonality=False)
model.add_seasonality(name='monthly', period=30.5, fourier_order=5)
model_fit = model.fit(prophet_df)
future = model.make_future_dataframe(periods=forecast_periods, freq=df.index.inferred_freq)
forecast_full = model_fit.predict(future)
forecast = forecast_full['yhat'].iloc[-forecast_periods:].values
# Add test data if available (only once to avoid clutter)
if test_size > 0 and idx == 0:
fig.add_trace(go.Scatter(x=df.index[train_size:], y=test_data, name='Test Data', line=dict(color='green')))
# Add forecast
label = f"Forecast Run {run_idx + 1}: {model_type}, {entry['train_percent']:.0f}/{entry['test_percent']:.0f}, {forecast_periods} periods"
fig.add_trace(go.Scatter(x=forecast_dates, y=forecast, name=label,
line=dict(dash='dash', color=colors[idx % len(colors)])))
fig.update_layout(title='Forecast Comparison', height=400, showlegend=True)
return pio.to_html(fig, full_html=False)