> ## Documentation Index
> Fetch the complete documentation index at: https://nixtla.io/docs/llms.txt
> Use this file to discover all available pages before exploring further.
# Improve Forecast Accuracy with TimeGPT
> Advanced techniques to enhance TimeGPT forecast accuracy for energy and electricity.
[](https://colab.research.google.com/github/Nixtla/nixtla/blob/main/nbs/docs/tutorials/22_how_to_improve_forecast_accuracy.ipynb)
# Improve Forecast Accuracy with TimeGPT
This guide demonstrates how to improve forecast accuracy using TimeGPT. We use hourly electricity price data from Germany as an illustrative example. Before you begin, make sure you have initialized the `NixtlaClient` object with your API key.
## Forecasting Results Overview
Below is a summary of our experiments and the corresponding accuracy improvements. We progressively refine forecasts by adding fine-tuning steps, adjusting loss functions, increasing the number of fine-tuned parameters, incorporating exogenous variables, and switching to a long-horizon model.
| Steps | Description | MAE | MAE Improvement (%) | RMSE | RMSE Improvement (%) |
| ----- | ---------------------------- | ---- | ------------------- | ---- | -------------------- |
| 0 | Zero-Shot TimeGPT | 18.5 | N/A | 20.0 | N/A |
| 1 | Add Fine-Tuning Steps | 11.5 | 38% | 12.6 | 37% |
| 2 | Adjust Fine-Tuning Loss | 9.6 | 48% | 11.0 | 45% |
| 3 | Fine-tune More Parameters | 9.0 | 51% | 11.3 | 44% |
| 4 | Add Exogenous Variables | 4.6 | 75% | 6.4 | 68% |
| 5 | Switch to Long-Horizon Model | 6.4 | 65% | 7.7 | 62% |
***
## Step-by-Step Guide
### Step 1: Install and Import Packages
Make sure all necessary libraries are installed and imported. Then set up the Nixtla client (replace with your actual API key).
```python theme={null}
import numpy as np
import pandas as pd
from utilsforecast.evaluation import evaluate
from utilsforecast.plotting import plot_series
from utilsforecast.losses import mae, rmse
from nixtla import NixtlaClient
nixtla_client = NixtlaClient(
# api_key='my_api_key_provided_by_nixtla'
)
```
### Step 2: Load the Dataset
We use hourly electricity price data from Germany (`unique_id == "DE"`). The final two days (`48` data points) form the test set.
```python theme={null}
df = pd.read_csv('https://raw.githubusercontent.com/Nixtla/transfer-learning-time-series/main/datasets/electricity-short-with-ex-vars.csv')
df['ds'] = pd.to_datetime(df['ds'])
df_sub = df.query('unique_id == "DE"')
df_train = df_sub.query('ds < "2017-12-29"')
df_test = df_sub.query('ds >= "2017-12-29"')
df_train.shape, df_test.shape
```
```bash Dataset Shape Output theme={null}
((1632, 12), (48, 12))
```
### Step 3: Benchmark Forecast with TimeGPT
**Info:** We first generate a zero-shot forecast using TimeGPT, which captures overall trends but may struggle with short-term fluctuations.
```python theme={null}
fcst_timegpt = nixtla_client.forecast(
df=df_train[['unique_id', 'ds', 'y']],
h=2*24,
target_col='y',
level=[90, 95]
)
```
```bash Forecast Logs theme={null}
[INFO logs here...]
```
#### Evaluation Metrics
| unique\_id | metric | TimeGPT |
| ---------- | ------ | ------- |
| DE | mae | 18.519 |
| DE | rmse | 20.038 |
### Step 4: Methods to Enhance Forecasting Accuracy
Use these following strategies to refine and improve your forecast:
#### 4.1 Add Fine-tuning Steps
Further fine-tuning typically reduces forecasting errors by adjusting the internal weights of the TimeGPT model, allowing it to better adapt to your specific data.
```python theme={null}
fcst_df = nixtla_client.forecast(df=df_train[['unique_id', 'ds', 'y']],
h=24*2,
finetune_steps = 30,
level=[90, 95])
```
Evaluation result:
| unique\_id | metric | TimeGPT |
| ---------- | ------ | ------- |
| DE | mae | 11.458 |
| DE | rmse | 12.643 |
#### 4.2 Fine-tune Using Different Loss Functions
Trying different loss functions (e.g., `MAE`, `MSE`) can yield better results for specific use cases.
```python theme={null}
fcst_df = nixtla_client.forecast(df=df_train[['unique_id', 'ds', 'y']],
h=24*2,
finetune_steps = 30,
finetune_loss = 'mae',
level=[90, 95])
```
Evaluation result:
| unique\_id | metric | TimeGPT |
| ---------- | ------ | ------- |
| DE | mae | 9.641 |
| DE | rmse | 10.956 |
#### 4.3 Adjust Number of Fine-tuned Parameters
The finetune\_depth parameter controls how many model layers are fine-tuned. It ranges from 1 (few parameters) to 5 (more parameters).
```python theme={null}
fcst_df = nixtla_client.forecast(df=df_train[['unique_id', 'ds', 'y']],
h=24*2,
finetune_steps = 30,
finetune_depth=2,
finetune_loss = 'mae',
level=[90, 95])
```
Evaluation result:
| unique\_id | metric | TimeGPT |
| ---------- | ------ | ------- |
| DE | mae | 9.002 |
| DE | rmse | 11.348 |
#### 4.4 Forecast with Exogenous Variables
Incorporate external data (e.g., weather conditions) to boost predictive performance.
```python theme={null}
#import exogenous variables
future_ex_vars_df = df_test.drop(columns = ['y'])
future_ex_vars_df.head()
#make forecast with historical and future exogenous variables
fcst_df = nixtla_client.forecast(df=df_train,
X_df=future_ex_vars_df,
h=24*2,
level=[90, 95])
```
Evaluation result:
| unique\_id | metric | TimeGPT |
| ---------- | ------ | ------- |
| DE | mae | 4.603 |
| DE | rmse | 6.359 |
#### 4.5 Use a Long-Horizon Model
For longer forecasting periods, models optimized for multi-step predictions tend to perform better. You can enable this by setting the model parameter to `timegpt-1-long-horizon`.
```python theme={null}
fcst_df = nixtla_client.forecast(df=df_train[['unique_id', 'ds', 'y']],
h=24*2,
model = 'timegpt-1-long-horizon',
level=[90, 95])
```
Evaluation result:
| unique\_id | metric | TimeGPT |
| ---------- | ------ | ------- |
| DE | mae | 6.366 |
| DE | rmse | 7.738 |
### Step 5: Conclusion and Next Steps
Key takeaways:
The following strategies offer consistent improvements in forecast accuracy. We recommend systematically experimenting with each approach to find the best combination for your data.
* Increase the number of fine-tuning steps.
* Experiment with different loss functions.
* Incorporate exogenous data.
* Switching to the long-horizon model for extended forecasting periods.
**Success:** Small refinements—like adding exogenous data or adjusting fine-tuning parameters—can significantly improve your forecasting results.
***
## Result Summary
| Steps | Description | MAE | MAE Improvement (%) | RMSE | RMSE Improvement (%) |
| ----- | ---------------------------- | ---- | ------------------- | ---- | -------------------- |
| 0 | Zero-Shot TimeGPT | 18.5 | N/A | 20.0 | N/A |
| 1 | Add Fine-Tuning Steps | 11.5 | 38% | 12.6 | 37% |
| 2 | Adjust Fine-Tuning Loss | 9.6 | 48% | 11.0 | 45% |
| 3 | Fine-tune More Parameters | 9.0 | 51% | 11.3 | 44% |
| 4 | Add Exogenous Variables | 4.6 | 75% | 6.4 | 68% |
| 5 | Switch to Long-Horizon Model | 6.4 | 65% | 7.7 | 62% |