Identifying model drift

Now you'll plot the model scores over time to visualize when drift occurs. By adding the threshold line and RMSE rolling windows, you can see how the trailing error lines indicate performance degradation.

The fc_log dataset with calculated moving averages, rmse_threshold, and Plotly as go have been pre-loaded for you.

This exercise is part of the course

Designing Forecasting Pipelines for Production

View Course

Hands-on interactive exercise

Have a go at this exercise by completing this sample code.

p = go.Figure()

# Add RMSE line
p.add_trace(go.Scatter(x=fc_log["forecast_start"], y=fc_log["____"],
                        mode='lines',
                        name='RMSE',
                        line=dict(color='royalblue', width=2)))

# Add the RMSE rolling windows for 7 and 14 days
p.add_trace(go.Scatter(x=fc_log["forecast_start"], y=fc_log["____"],
                        mode='lines',
                        name='7 Days MA',
                        line=dict(color='green', width=2)))

p.add_trace(go.Scatter(x=fc_log["forecast_start"], y=fc_log["____"],
                        mode='lines',
                        name='14 Days MA',
                        line=dict(color='orange', width=2)))

# Add the threshold line
p.add_trace(go.Scatter(x=[fc_log["forecast_start"].min(), fc_log["forecast_start"].max()], 
y=[rmse_threshold, rmse_threshold], 
name="Threshold",
line=dict(color="red", width=2, dash="dash")))

# Add plot titles and show the plot
p.update_layout(title="Forecast Error Rate Over Time",
                xaxis_title="____",
                yaxis_title="____", 
                height=400,
                title_x=0.5,
                margin=dict(t=50, b=50, l=50, r=50))
p.show()

Edit and Run Code