README.md

metadata

title: DATFID MASTER
emoji: 🤖
colorFrom: indigo
colorTo: pink
sdk: docker
pinned: false
short_description: DATFID - Secure and powerful forecasting SDK

DATFID MASTER

DATFID delivers reliable and interpretable forecasting for any time series. Our models consistently rank at the top—see our website datfid.com for live results.

We achieved state-of-the-art accuracy in the M5 Forecasting Competition, beated all the benchmarks, forecasting hierarchical Walmart product sales over a 5-year period using advanced exogenous features and rigorous error metrics (Kaggle M5 page).

---

Why DATFID Stands Out

• Professionally recognized benchmarking
  M5 is part of the renowned “M-Competitions” series, used worldwide to compare forecasting methods under realistic conditions. (Makridakis Competitions, Wikipedia).

• Hierarchical & intermittent series
  DATFID handles complex retail data with product-store hierarchies, zero-heavy series, pricing, calendar events, and promotional features—all core challenges of the M5 setup (Background & Organization of the M5 Competition).

• Robust accuracy
  Scoring well on 42,840 series across multiple aggregation levels is a testament to our method’s strength—especially on challenging intermittent product-level data.

• Fast, scalable, and interpretable
  Unlike “black-box” approaches, we prioritize transparency and efficiency without sacrificing accuracy.

---

Getting Started

1. Request a Token

To access the API, request your personal DATFID token via admin@datfid.com.

2. Install the SDK

pip install -i https://test.pypi.org/simple/ datfid

3. Example Usage

import pandas as pd
from datfid.client import DATFIDClient

# Initialize with your DATFID token
client = DATFIDClient(token="dt+YOUR_TOKEN_HERE")

# Fit the model
fit_result = client.fit_model(
    df=my_dataframe,
    id_col="SKU_ID",
    time_col="Date",
    y="Sales",
    lagged_features={"Sales": 2, "Promo": 1},
    current_features=["Price", "Holiday"],
    filter_by_significance=True,
    meanvar_test=True
)

# Forecast
forecast_df = client.forecast_model(df_forecast=my_forecast_dataframe)
print(forecast_df.head())

# Check token validity
client.secure_ping()

---

Contact & Info

• Website: datfid.com
• Get your token: admin@datfid.com

---

Happy forecasting with confidence!
— The DATFID Team