DATFID SDK

A Python SDK to access the DATFID API to forecast your data.

Features

• Easy model fitting: Build panel data models with time-dependent and static features.
• Flexible lag handling: Specify lags for the dependent variable and selected features.
• Forecasting: Generate future predictions with aligned timestamps and IDs.
• Statistical options: Filter features by significance and apply mean-variance tests.
• White box full interpretability: Get fully interpretable model with equation, estimated parameters, and standard errors.

Installation

pip install datfid

Usage

Before using the SDK, please request an access token by emailing admin@datfid.com or by visiting our website datfid.com.

from datfid import DATFIDClient

# Initialize the client with your DATFID token
client = DATFIDClient(token="your_DATFID_token")

# Fit a model
fit_result = client.fit_model(
    df=dataframe,
    id_col="name of id column",
    time_col="name of time column",
    y="name of dependent variable",
    lag_y="starting lag : ending lag",
    lagged_features={
        "feature 1": "starting lag : ending lag",
        "feature 2": "starting lag : ending lag"
    },
    current_features=["feature 3", "feature 4"],
    filter_by_significance=True/False,
    dummy_sqrt=True/False,
    dummy_interact=True/False,
    dummy_logs=True/False,
    dummy_yejo=True/False,
    dummy_asinh=True/False,
    dummy_logs_interact=True/False,
    meanvar_test=True/False
)

# Generate forecasts
forecast_df = client.forecast_model(
    df_forecast=dataframe
)

# The forecast DataFrame contains the individual IDs and timestamps
# from the original data plus a "forecast" column with predicted values.

Example 1

Sample dataset from GitHub (Food and Beverages demand forecasting):

import pandas as pd
from datfid import DATFIDClient

# Initialize the client with your DATFID token
client = DATFIDClient(token="your_DATFID_token")

# Load dataset for model fitting
url_fit = "https://raw.githubusercontent.com/datfid-valeriidashuk/sample-datasets/main/Food_Beverages.xlsx"
df = pd.read_excel(url_fit)

# Fit the model
result = client.fit_model(df=df,
                          id_col="Product",
                          time_col="Time",
                          y="Revenue",
                          current_features='all',
                          filter_by_significance=True
                          )

# Load dataset for forecasting
url_forecast = "https://raw.githubusercontent.com/datfid-valeriidashuk/sample-datasets/main/Food_Beverages_forecast.xlsx"
df_forecast = pd.read_excel(url_forecast)

# Forecast revenue using the fitted model
forecast = client.forecast_model(df_forecast=df_forecast)

Example 2

Slightly larger sample dataset from GitHub (Banking sector, forecasting loan probability):

import pandas as pd
from datfid import DATFIDClient

# Initialize the client with your DATFID token
client = DATFIDClient(token="your_DATFID_token")

# Load dataset for model fitting
url_fit = "https://raw.githubusercontent.com/datfid-valeriidashuk/sample-datasets/main/Banking_extended.xlsx"
df = pd.read_excel(url_fit)

# Fit the model
result = client.fit_model(df=df,
                          id_col="Individual",
                          time_col="Time",
                          y="Loan Probability",
                          lag_y="1:3",
                          lagged_features={"Income Level": "1:3"},
                          filter_by_significance=True)

# Load dataset for forecasting
url_forecast = "https://raw.githubusercontent.com/datfid-valeriidashuk/sample-datasets/main/Banking_extended_forecast.xlsx"
df_forecast = pd.read_excel(url_forecast)

# Forecast loan probability using the fitted model
forecast = client.forecast_model(df_forecast=df_forecast)

API Reference

DATFIDClient

client = DATFIDClient(token: str)

Initialize the client with your DATFID token.

client.fit_model(df: pd.DataFrame, id_col: str, time_col: str, y: str, lag_y: Optional[Union[int, str, list[int]]] = None, lagged_features: Optional[Dict[str, int]] = None, current_features: Optional[list] = None, filter_by_significance: bool = False, meanvar_test: bool = False) -> SimpleNamespace

Fit a model using the provided dataset.

client.forecast_model(df_forecast: pd.DataFrame) -> pd.DataFrame

Generate forecasts using the fitted model.