45 lines
1.6 KiB
Python
45 lines
1.6 KiB
Python
#basic imports
|
|
import pandas
|
|
from sklearn.linear_model import LinearRegression
|
|
|
|
#basic paths
|
|
openTrain = './train/train.tsv'
|
|
openDev = './dev-0/in.tsv'
|
|
openTest = './test-A/in.tsv'
|
|
|
|
#read from files
|
|
with open('./names') as f_names:
|
|
names = f_names.read().rstrip('\n').split('\t')
|
|
read0 = pandas.read_table(openTrain, sep='\t', names=names)
|
|
read1 = pandas.read_table(openDev, sep='\t', names=['mileage', 'year', 'brand', 'engineType', 'engineCapacity'])
|
|
|
|
#basic normalization & filtering
|
|
trainSet = pandas.get_dummies(read0, columns=['engineType'])
|
|
categories1 = trainSet.select_dtypes(include=object).columns.values
|
|
for c in categories1:
|
|
trainSet[c] = trainSet[c].astype('category').cat.codes
|
|
trainSet = trainSet.loc[(trainSet['price'] > 10000)] #to avoid suspicious observations
|
|
#for some reason this value gives the smallest RMSE according to geval, while smaller or bigger
|
|
#price gives RMSE >34k
|
|
|
|
#Model training
|
|
X = trainSet.loc[:, trainSet.columns != 'price']
|
|
solution = LinearRegression().fit(X, trainSet['price'])
|
|
|
|
devSet = pandas.get_dummies(read1, columns=['engineType'])
|
|
categories2 = devSet.select_dtypes(include=object).columns.values
|
|
for c in categories2:
|
|
devSet[c] = devSet[c].astype('category').cat.codes
|
|
|
|
predict = solution.predict(devSet)
|
|
predict.tofile("./dev-0/out.tsv", sep='\n')
|
|
testSet = pandas.get_dummies(read1, columns=['engineType'])
|
|
|
|
categories3 = testSet.select_dtypes(include=object).columns.values
|
|
for c in categories3:
|
|
testSet[c] = testSet[c].astype('category').cat.codes
|
|
|
|
predict = solution.predict(devSet)
|
|
predict.tofile("./test-A/out.tsv", sep='\n')
|
|
|
|
#Outcome: 33956 for prices >10000 |