# shenweichen/DeepCTR

Easy-to-use,Modular and Extendible package of deep-learning based CTR models .

Repository: https://github.com/shenweichen/DeepCTR
Canonical: https://ross.abutalabs.com/products/deepctr
Homepage: https://deepctr-doc.readthedocs.io/en/latest/index.html
Language: Python
License: Apache-2.0
License Family: permissive
Topics: ctr, click-through-rate, deep-learning, factorization-machines, deepfm, ffm, nfm, mlr, din, deepinterestnetwork, xdeepfm, deepcross, autoint, deepinterestevolutionnetwork, dien, fgcnn, recommendation, mmoe, esmm, ple
Last push: 2026-07-02T18:21:47+00:00

## Health v2 (maintenance only)
Score: 77/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 90, release rhythm 48, longevity 100
- inputs: {"age_days": 3252, "days_push": 62, "days_rel": 139, "gap_med": null, "n_releases_24m": 1}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 8050, forks 2219 (observed 2026-08-28T04:10:12.297311+00:00)

## What it is
DeepCTR is a Python library of easy-to-use, modular, and extendible deep-learning based CTR (click-through rate) prediction models built on TensorFlow 1.15 and 2.x. It provides tf.keras-like fit/predict interfaces plus a TensorFlow Estimator interface for large-scale distributed training, along with reusable layers for building custom models.

## Use cases
- train a DeepFM model for click-through rate prediction
- build custom CTR models from reusable layers
- run distributed CTR training on large datasets with TFRecord and Estimator
- experiment with models like DIN, xDeepFM, AutoInt, and DIEN
- do multi-task learning with MMOE, ESMM, or PLE
- benchmark deep CTR models on the Criteo dataset

## When to choose
- you need production-ready deep CTR models in TensorFlow with a familiar keras-style API
- you want many state-of-the-art CTR and multi-task models behind one consistent interface
- you need both quick experimentation and large-scale distributed training paths

## When to avoid
- your stack is PyTorch (use DeepCTR-Torch instead)
- you need general recommendation/retrieval models rather than CTR prediction (consider DeepMatch)
- you want a framework-agnostic or non-TensorFlow solution

## Facets
- artifact type: library
- maturity: stable
- function: machine-learning, deep-learning
- domain: machine-learning, deep-learning, data-science
- platform: python
- tags: ctr, click-through-rate, recommendation, tensorflow, keras, factorization-machines, deepfm, din, multi-task-learning, gpu

## Member repositories
- shenweichen/DeepCTR (main) score 77

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:10:12.297311+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T17:31:24.338215+00:00, confidence not recorded.
  - readme: https://github.com/shenweichen/DeepCTR (fetched 2026-08-28T04:10:12.297311+00:00, sha 1cee60a0879a)
  - homepage: https://deepctr-doc.readthedocs.io/en/latest/index.html (fetched 2026-08-29T08:29:03.752552+00:00, sha a05afc4e753d)
  - registry_pypi: https://pypi.org/pypi/deepctr/json (fetched 2026-08-29T08:29:03.761737+00:00, sha d860c4605ab3)
- Data as of 2026-08-30T08:39:29.467469+00:00.