2021 credit card consuming recommendation

Overview

2021-credit-card-consuming-recommendation

My implementation and sharing of this contest: https://tbrain.trendmicro.com.tw/Competitions/Details/18. I got rank 9 in the Private Leaderboard.

Run My Implementation

Required libs

matplotlib, numpy, pytorch, and yaml. Versions of them are not restricted as long as they're new enough.

Preprocess

python3 data_to_pkl.py
  • The officially provided csv file should be in data dir.
  • Output pkl file is also in data dir.

Feature Extraction

python3 pkl_to_fea_allow_shorter.py
  • See "作法分享" for detailed description of optional parameters.

Training

python3 train_cv_allow_shorter.py -s save_model_dir
  • -s: where you want to save the trained model.

Inference

Generate model outputs

python3 test_cv_raw_allow_shorter.py model_dir max_len
  • model_dir: directory of the trained model.
  • max_len: max number of month considered for each customer.

Merge model outputs

python3 test_cv_merge_allow_shorter.py n_fold_train
  • n_fold_train: number of folds used for training.

作法分享

以下將介紹本競賽所使用的執行環境、特徵截取、模型設計與訓練。

執行環境

硬體方面,初始時使用 ASUS P2440 UF 筆電,含 i7-8550U CPU 及 MX130 顯示卡,主記憶體擴充至 20 GB;後續使用較多特徵及較長期間的資料時,改為使用 AWS p2.xlarge 機器,含 K80 顯示卡以及約 64 GB 主記憶體。AWS 的經費來源是上一個比賽進入複賽拿到的點數,在打完複賽後還有剩下來的部分。

程式語言為 Python 3,未特別指定版本;函式庫則如本說明前半部所示,其中的 matplotlib 為繪圖觀察用,而 yaml 為儲存模型組態用。

特徵截取(附帶資料觀察)與預測目標

我先將欄位分為兩類,依照「訓練資料欄位說明」的順序,從 shop_tag(消費類別)起至 card_other_txn_amt_pct (其他卡片消費金額佔比)止,因為是從每月每類的消費行為而來,且消費行為必然是變動的,因此列為「時間變化類」;而 masts (婚姻狀態)起至最後為止,因所觀察到的每人的婚姻狀態或教育程度等,在比賽資料所截取的兩年間幾乎都不會變化,故列為「時間不變類」,以節省運算及儲存資源。事實上,在「時間不變類」的欄位當中,平均每人用過的不同狀態,平均約為 1.005 至 1.167 種,最多的則為 3 至 5 種。

時間變化類

對於每人每月的消費紀錄,以如下步驟取特徵

  1. 排序出消費金額前 n 大者,最佳成績中使用的 n 為 13。根據觀察,約 99% 的人,其每月消費類別數在 13 以下。
  2. 取該月時間特徵,為待預測月減去該月,共 1 維。
  3. 該月類別特徵共 49 維,若該月該類別消費金額在該月前 n 名中且金額大於 0 者,其特徵值由名次大到小依次為 n, n-1, n-2, …, 1;前 n 名以外或金額小於等於 0 的類別,其特徵值為 0。
  4. 對於前 n 名的每個類別,無論其消費金額皆取以下特徵,共 22 維:txn_cnt, txn_amt, domestic_offline_cnt, domestic_online_cnt, overseas_offline_cnt, overseas_online_cnt, domestic_offline_amt_pct, domestic_online_amt_pct, overseas_offline_amt_pct, overseas_online_amt_pct, card_*_txn_cnt (* = 1, 2, 4, 6, 10, other), card_*_txn_amt_pct (* = 1, 2, 4, 6, 10, other)。
    • 1, 2, 4, 6, 10, other 為所有消費紀錄中,使用次數最多的前六個卡片編號。
  5. 以上共 1 + 49 + 13 * 22 = 336 維

跨月份的取值方式如下圖所示,其中每個圓角方塊代表每人的一個月份的所有消費紀錄,而 N1 為 20 個月,N2 為 4 組,在範圍內會盡可能的取長或多。另,若該月未有消費紀錄,則忽略該月。

時間變化類取值方式

時間不變類

對於每位客戶,僅使用取值範圍內最後消費當月(N1 範圍內的最後一筆)的金額最大的類別所記載的資料來組成特徵。

使用時,以 masts, gender_code, age, primary_card, slam 各自編成 one-hot encoding 或數值型態後組合,共得 20 維,細節說明如下

  • masts: 含缺值共 4 種狀態,4 維。
  • gender_code: 含缺值共 3 種狀態,3 維。
  • age: 含缺值共 10 種狀態,10 維。
  • primary_card: 沒有缺值,共 2 種狀態,2 維。
  • slam: 數值型態,取 log 後做為特徵,1 維。

此部分亦嘗試過其他特徵,但可能是因為維度較大不易訓練(如 cuorg,含缺值共 35 維),或客戶有可能填寫不實(如 poscd),故未取得較好之結果。

預測目標

共 16 維,代表需要預測的 16 個類別,其中下月金額第一名者為 1,第二名者 0.8,第三名者 0.6,第四名以下有購買者 0.2,未購買者 0。

小結

以上取法經去除輸出全部為 0 (即預測目標月份沒有購買行為)之資料後,共約 102 萬組。

模型設計與訓練

本次比賽使用的模型架構如下圖,主體為 BiLSTM + attention,前後加上適量的 linear layers,其中標色部分為 attention 的做用範圍,最後面的 dense layers 之細部架構則為 (dense 128 + ReLU + dropout 0.1) * 2 + dense 16 + Sigmoid。

模型架構

訓練方式為 5 folds cross validation,預測時會將五個模型的結果取平均,再依據平均後的排名輸出前三名的類別。細節參數如下,未提及之參數係依照 pytorch 預設值,未進行修改:

  • Num of epochs: 100 epochs,若 validation loss 連續 10 個 epochs 未創新低,則提前終止該 fold 的訓練。
  • Batch size: 512。
  • Loss: MSE。
  • Optimizer: ADAM with learning rate 0.01。
  • Learning rate scheduler: 每個 epoch 下降為上一次的 0.95 倍,直至其低於 0.0001 為止。
Owner
Wang, Chung-Che
Wang, Chung-Che
Easily pull telemetry data and create beautiful visualizations for analysis.

This repository is a work in progress. Anything and everything is subject to change. Porpo Table of Contents Porpo Table of Contents General Informati

Ryan Dawes 33 Nov 30, 2022
Selecting Parallel In-domain Sentences for Neural Machine Translation Using Monolingual Texts

DataSelection-NMT Selecting Parallel In-domain Sentences for Neural Machine Translation Using Monolingual Texts Quick update: The paper got accepted o

Javad Pourmostafa 6 Jan 07, 2023
Code for the paper "TadGAN: Time Series Anomaly Detection Using Generative Adversarial Networks"

TadGAN: Time Series Anomaly Detection Using Generative Adversarial Networks This is a Python3 / Pytorch implementation of TadGAN paper. The associated

Arun 92 Dec 03, 2022
CLIPImageClassifier wraps clip image model from transformers

CLIPImageClassifier CLIPImageClassifier wraps clip image model from transformers. CLIPImageClassifier is initialized with the argument classes, these

Jina AI 6 Sep 12, 2022
Christmas face app for Decathlon xmas coding party!

Christmas Face Application Use this library to create the perfect picture for your christmas cards! Done by Hasib Zunair, Guillaume Brassard and Samue

Hasib Zunair 4 Dec 20, 2021
NALSM: Neuron-Astrocyte Liquid State Machine

NALSM: Neuron-Astrocyte Liquid State Machine This package is a Tensorflow implementation of the Neuron-Astrocyte Liquid State Machine (NALSM) that int

Computational Brain Lab 4 Nov 28, 2022
Neural machine translation between the writings of Shakespeare and modern English using TensorFlow

Shakespeare translations using TensorFlow This is an example of using the new Google's TensorFlow library on monolingual translation going from modern

Motoki Wu 245 Dec 28, 2022
Codebase for testing whether hidden states of neural networks encode discrete structures.

structural-probes Codebase for testing whether hidden states of neural networks encode discrete structures. Based on the paper A Structural Probe for

John Hewitt 349 Dec 17, 2022
How to Learn a Domain Adaptive Event Simulator? ACM MM, 2021

LETGAN How to Learn a Domain Adaptive Event Simulator? ACM MM 2021 Running Environment: pytorch=1.4, 1 NVIDIA-1080TI. More details can be found in pap

CVTEAM 4 Sep 20, 2022
U-2-Net: U Square Net - Modified for paired image training of style transfer

U2-Net: U Square Net Modified for paired image training of style transfer This is an unofficial repo making use of the code which was made available b

Doron Adler 43 Oct 03, 2022
Bottom-up attention model for image captioning and VQA, based on Faster R-CNN and Visual Genome

bottom-up-attention This code implements a bottom-up attention model, based on multi-gpu training of Faster R-CNN with ResNet-101, using object and at

Peter Anderson 1.3k Jan 09, 2023
A pytorch reproduction of { Co-occurrence Feature Learning from Skeleton Data for Action Recognition and Detection with Hierarchical Aggregation }.

A PyTorch Reproduction of HCN Co-occurrence Feature Learning from Skeleton Data for Action Recognition and Detection with Hierarchical Aggregation. Ch

Guyue Hu 210 Dec 31, 2022
Official PyTorch code for the paper: "Point-Based Modeling of Human Clothing" (ICCV 2021)

Point-Based Modeling of Human Clothing Paper | Project page | Video This is an official PyTorch code repository of the paper "Point-Based Modeling of

Visual Understanding Lab @ Samsung AI Center Moscow 64 Nov 22, 2022
The BCNet related data and inference model.

BCNet This repository includes the some source code and related dataset of paper BCNet: Learning Body and Cloth Shape from A Single Image, ECCV 2020,

81 Dec 12, 2022
Make your master artistic punk avatar through machine learning world famous paintings.

Master-art-punk Make your master artistic punk avatar through machine learning world famous paintings. 通过机器学习世界名画制作属于你的大师级艺术朋克头像 Nowadays, NFT is beco

Philipjhc 53 Dec 27, 2022
Leaderboard and Visualization for RLCard

RLCard Showdown This is the GUI support for the RLCard project and DouZero project. RLCard-Showdown provides evaluation and visualization tools to hel

Data Analytics Lab at Texas A&M University 246 Dec 26, 2022
An easy way to build PyTorch datasets. Modularly build datasets and automatically cache processed results

EasyDatas An easy way to build PyTorch datasets. Modularly build datasets and automatically cache processed results Installation pip install git+https

Ximing Yang 4 Dec 14, 2021
Text Extraction Formulation + Feedback Loop for state-of-the-art WSD (EMNLP 2021)

ConSeC is a novel approach to Word Sense Disambiguation (WSD), accepted at EMNLP 2021. It frames WSD as a text extraction task and features a feedback loop strategy that allows the disambiguation of

Sapienza NLP group 36 Dec 13, 2022
This is the official code release for the paper Shape and Material Capture at Home

This is the official code release for the paper Shape and Material Capture at Home. The code enables you to reconstruct a 3D mesh and Cook-Torrance BRDF from one or more images captured with a flashl

89 Dec 10, 2022