All Projects → Hanszhuang → mobileRiskUser

Hanszhuang / mobileRiskUser

Licence: other
基于移动网络通讯行为的风险用户识别 (15th/624)

Programming Languages

Jupyter Notebook
11667 projects

Projects that are alternatives of or similar to mobileRiskUser

ml-time-series-analysis-on-sales-data
Time Series Decomposition techniques and random forest algorithm on sales data
Stars: ✭ 34 (+17.24%)
Mutual labels:  datamining
Kaggle-Competition-Sberbank
Top 1% rankings (22/3270) code sharing for Kaggle competition Sberbank Russian Housing Market: https://www.kaggle.com/c/sberbank-russian-housing-market
Stars: ✭ 31 (+6.9%)
Mutual labels:  lightgbm
ai-deployment
关注AI模型上线、模型部署
Stars: ✭ 149 (+413.79%)
Mutual labels:  lightgbm
unleash-client-java
Unleash client SDK for Java
Stars: ✭ 86 (+196.55%)
Mutual labels:  feature
kaggle-plasticc
Solution to Kaggle's PLAsTiCC Astronomical Classification Competition
Stars: ✭ 50 (+72.41%)
Mutual labels:  lightgbm
Tencent2017 Final Rank28 code
2017第一届腾讯社交广告高校算法大赛Rank28_code
Stars: ✭ 85 (+193.1%)
Mutual labels:  lightgbm
SynapseML
Simple and Distributed Machine Learning
Stars: ✭ 3,355 (+11468.97%)
Mutual labels:  lightgbm
Arch-Data-Science
Archlinux PKGBUILDs for Data Science, Machine Learning, Deep Learning, NLP and Computer Vision
Stars: ✭ 92 (+217.24%)
Mutual labels:  lightgbm
ETL-Starter-Kit
📁 Extract, Transform, Load (ETL) 👷 refers to a process in database usage and especially in data warehousing. This repository contains a starter kit featuring ETL related work.
Stars: ✭ 21 (-27.59%)
Mutual labels:  datamining
jest-launchdarkly-mock
Easily unit test LaunchDarkly feature flagged components with jest
Stars: ✭ 14 (-51.72%)
Mutual labels:  feature
autogbt-alt
An experimental Python package that reimplements AutoGBT using LightGBM and Optuna.
Stars: ✭ 76 (+162.07%)
Mutual labels:  lightgbm
DeepCD
[ICCV17] DeepCD: Learning Deep Complementary Descriptors for Patch Representations
Stars: ✭ 39 (+34.48%)
Mutual labels:  feature
xgboost-lightgbm-hyperparameter-tuning
Bayesian Optimization and Grid Search for xgboost/lightgbm
Stars: ✭ 40 (+37.93%)
Mutual labels:  lightgbm
recsys2019
The complete code and notebooks used for the ACM Recommender Systems Challenge 2019
Stars: ✭ 26 (-10.34%)
Mutual labels:  lightgbm
scikit-mine
scikit-mine : pattern mining in Python
Stars: ✭ 45 (+55.17%)
Mutual labels:  datamining
aws-machine-learning-university-dte
Machine Learning University: Decision Trees and Ensemble Methods
Stars: ✭ 119 (+310.34%)
Mutual labels:  lightgbm
Friends-Recommender-In-Social-Network
Friends Recommendation and Link Prediction in Social Netowork
Stars: ✭ 33 (+13.79%)
Mutual labels:  datamining
HumanOrRobot
a solution for competition of kaggle `Human or Robot`
Stars: ✭ 16 (-44.83%)
Mutual labels:  lightgbm
arfs
All Relevant Feature Selection
Stars: ✭ 21 (-27.59%)
Mutual labels:  lightgbm
lleaves
Compiler for LightGBM gradient-boosted trees, based on LLVM. Speeds up prediction by ≥10x.
Stars: ✭ 132 (+355.17%)
Mutual labels:  lightgbm

基于移动网络通讯行为的风险用户识别

######## 1. 2018-05-22 baseline

因为这学期担任实训课TA,这个比赛是实训课的作业,所以做个baseline给师弟师妹们参考,这个baseline 线下是0.80,线上成绩是0.76~0.77
(所以LB上有一群DM+学号的,不要惊讶,不是小号,是一群可爱的小鲜肉)

######## 2. 2018-06-12 解决方案 (初赛 3th, 复赛 15th)

特征工程:

2.1 单变量数量统计特征:

voice统计用户 记录数,用户 不同opp_num 记录数

voice统计用户不同 opp_head记录 数

voice统计用户 不同 opp_len 的记录数

统计用户不同 call_tyoe 记录数

sms统计用户sms 不同opp_num记录数

sms统计用户sms 不同opp_head 记录数

sms统计用户 不同opp_len记录数

sms不同in_out 记录数

2.2 多变量数目统计特征:

voice统计用户不同 in_out 下 不同 opp_num记录数

voice统计不同 opp_len 下 不同 opp_head 记录数

voice统计不同 opp_len 下不同opp_head 的记录数

voice统计不同call_type 下 不同opp_num 的记录数

sms统计用户不同opp_len 中 不同opp_head的记录数

sms不同in_out 下 不同opp_head 数

2.3 one-hot 类数目统计特征:

voice对opp_num one-hot 统计记录数

sms每天不同的opp_head 记录数

sms opp_head one-hot 统计记录数

sms每天的短信记录数

wa top 1000 wa_name 分组统计记录数

2.4 时间统计量:

voice通话时长统计量

voice两次通话间隔统计量

sms两次短信间隔统计量

模型:

只使用了lgb单模型,成绩是初赛第三(0.874),复赛14(0.864),应该是wa 特征 A,B榜数据分布不一致。

最近一堆事,趁着早上上班前,赶紧把方案开源了。

Note that the project description data, including the texts, logos, images, and/or trademarks, for each open source project belongs to its rightful owner. If you wish to add or remove any projects, please contact us at [email protected].