The latest data mining scheme sort out!

There are 1500 words in this article, it is recommended to read for 9 minutes

This article takes you immersive to experience the participation process from 0 to 1 in the prediction of diabetic genetic risk prediction.

Introduction to the topic

HKUST Xunfei: Diabetes Genetic Risk Test Challenge. Background: As of 2022, there are nearly 130 million diabetic patients in China. Causes of Chinese diabetes are affected by various factors such as lifestyle, aging, urbanization, and family genetic. At the same time, diabetic patients tend to be younger.

Diabetes can cause complications of cardiovascular, kidney, and cerebrovascular complications. Therefore, it has very important clinical significance to accurately diagnose individuals with diabetes. Early genetic risk prediction of diabetes will help prevent the occurrence of diabetes.

The address of the event:

http://challenge.xfyun.cn/topic/info?Type=diabetesch=ds22-dw-gzy01

Source code:

https://github.com/datawhalechina/competition-baseline

Tie task

In this competition, you need to build a diabetic genetic risk prediction model through training data sets, and then predict whether the individual of the test data set has diabetes, and help diabetic patients to solve this "sweet trouble" with us.

Tournament data

The topic data is composed of training sets and test collection. The specific situation is as follows:

Training set: There are 5070 pieces of data, which is used to build your prediction model

Test set: There are 1,000 pieces of data for verifying the performance of the predicted model.

Among them, the training collection data includes 9 fields: gender, year of birth, weight index, family history of diabetes, diastolic pressure, oral sugar resistance test, insulin release experiments, triceis muscle fold thickness, diabetes signs (data labels) Essence

Grading

The F1-SCORE indicator in the binary class task is evaluated. The larger the F1-SCORE shows that the better the predicted model, the better the F1-SCORE definition is as follows:

in:

Treatment Baseline import data

Import Pandas as PD Import LightGBM

Pre -processing

data1 = pd.read_csv ('' competition training set .csv ', encoding =' gbk ') data2 = pd.read_csv (' 'competition test set .csv', enCoding = 'gbk') #Label marked as -1 data2 ['Diabetes logo'] =-1 #<<< data = pd.concat ([data1, data2], axis = 0, iGnore_index = true) characteristics "" "" "" "" "" "" "" " The normal value of the human body weight index of the human body is between 18.5-24 below 18.5 is too light weight index between 24-27 is Overweight 27 or above is considered obesity higher than 32 is very obese.

not > Elif 18.5 <= A <= 24: Return 1 Elif 24 Return 2 Elif 27 Return 3 Else: Return 4 Data ['BMI'] = Data ['Weight Index'] .apply (BMI) Data ['Birth Year'] = 2022-Data ['Birth Year'] #Stead to age #Diabetes Family History "" " No record Uncle or aunt has one with diabetes/uncle or aunt with diabetes with diabetes One of the parents suffering from diabetes "" " defff (A): if A == 'No record': Return 0 Elif A == 'Uncle or Auntie has a diabetes' or a ==' Uncle or Auntie. Diabetes ': Return 1 Else: Return 2 Data [' Diabetes Family History '] = Data [' Diabetes Family History '] .apply (FHOD) Data [' diastolic pressure '] = data [' diastolic pressure '] .Fillna (-1) CODE> "" " The range of diastolic blood pressure is 60-90 " "" Def Dbp (a): IF a <60: Return 0 Elif 60 <= A Return 1 Elif A & 90: Return 2 Else: Return a data ['dbp' ] = Data ['Solid Pressure'] .apply (DBP) Data Training Data/Test Data Preparation