0
$\begingroup$

I have some dataset which contains different paramteres and data.head() looks like this

enter image description here

Applied some preprocessing and performed Feature ranking -

dataset = pd.read_csv("ML.csv",header = 0)

#Get dataset breif
print(dataset.shape)
print(dataset.isnull().sum())
#print(dataset.head())

#Data Pre-processing
data = dataset.drop('organization_id',1)
data = data.drop('status',1)
data = data.drop('city',1)

#Find median for features having NaN
median_zip, median_role_id, median_specialty_id, median_latitude, median_longitude = data['zip'].median(),data['role_id'].median(),data['specialty_id'].median(),data['latitude'].median(),data['longitude'].median() 
data['zip'].fillna(median_zip, inplace=True)
data['role_id'].fillna(median_role_id, inplace=True)
data['specialty_id'].fillna(median_specialty_id, inplace=True)
data['latitude'].fillna(median_latitude, inplace=True)
data['longitude'].fillna(median_longitude, inplace=True)

#Fill YearOFExp with 0
data['years_of_experience'].fillna(0, inplace=True)
target = dataset.location_id

#Perform Recursive Feature Extraction
svm = LinearSVC()
rfe = RFE(svm, 1)
rfe = rfe.fit(data, target) #IT give convergence Warning - Normally when an optimization algorithm does not converge, it is usually because the problem is not well-conditioned, perhaps due to a poor scaling of the decision variables. 


names = list(data)
print("Features sorted by their score:")
print(sorted(zip(map(lambda x: round(x, 4), rfe.ranking_), names)))

Output

Features sorted by their score:

[(1, 'location_id'), (2, 'department_id'), (3, 'latitude'), (4, 'specialty_id'), (5, 'longitude'), (6, 'zip'), (7, 'shift_id'), (8, 'user_id'), (9, 'role_id'), (10, 'open_positions'), (11, 'years_of_experience')]

From this I understand that which parameters have more importance. Is above processing correct to understand the feature important. How can I use above information for better model training?

When I to model training it gives very high accuracy. How come it gives so high accuracy?

from sklearn.linear_model import LinearRegression
from sklearn.model_selection import train_test_split

dataset = pd.read_csv("prod_data_for_ML.csv",header = 0)

#Data Pre-processing
data = dataset.drop('location_id',1)
data = data.drop('status',1)
data = data.drop('city',1)

#Find median for features having NaN
median_zip, median_role_id, median_specialty_id, median_latitude, median_longitude = data['zip'].median(),data['role_id'].median(),data['specialty_id'].median(),data['latitude'].median(),data['longitude'].median() 
data['zip'].fillna(median_zip, inplace=True)
data['role_id'].fillna(median_role_id, inplace=True)
data['specialty_id'].fillna(median_specialty_id, inplace=True)
data['latitude'].fillna(median_latitude, inplace=True)
data['longitude'].fillna(median_longitude, inplace=True)

#Fill YearOFExp with 0
data['years_of_experience'].fillna(0, inplace=True)

#Start training

labels = dataset.location_id
train1 = data
algo = LinearRegression()
x_train , x_test , y_train , y_test = train_test_split(train1 , labels , test_size = 0.20,random_state =1)

# x_train.to_csv("x_train.csv", sep=',', encoding='utf-8')
# x_test.to_csv("x_test.csv", sep=',', encoding='utf-8')

algo.fit(x_train,y_train)
algo.score(x_test,y_test)

output

0.981150074104111

from sklearn import ensemble
clf = ensemble.GradientBoostingRegressor(n_estimators = 400, max_depth = 5, min_samples_split = 2,
          learning_rate = 0.1, loss = 'ls')
clf.fit(x_train, y_train)
clf.score(x_test,y_test)

Output -

0.99

What I want to do is, predicting location-id correctly. Am I doing anything wrong? What ithe s correct way to build model for this sort of situati?n.

I know there is some way that I can get Precision, recall, f1 for each paramteres. Can anyone give me reference link to perform this.strong text

$\endgroup$
  • $\begingroup$ Are you trying to predict location_id ? $\endgroup$ – Shamit Verma Mar 15 at 14:06
  • $\begingroup$ @ShamitVerma: Yes $\endgroup$ – Jhon Patric Mar 15 at 14:13
0
$\begingroup$

Train data includes latitude, longitude and zipcode. Output variable is location.

It is trivial to predict location if zip and lat,lon are known. Try removing these attributes and see if that has any impact on validation score.

$\endgroup$
  • $\begingroup$ Meaningwise location ID is related to lat, long and zip. But I guess location_ID is independent of latitude, longitude and zipcode when we use it for recommendation. I am not sure. But let me try it $\endgroup$ – Jhon Patric Mar 15 at 14:15
  • $\begingroup$ Also, though I have considered those parameters, prediction score is very high. If it create negative impact then it should degrade the prediction score. If my understanding is not wrong $\endgroup$ – Jhon Patric Mar 15 at 14:16
  • $\begingroup$ can a location_id have different lat,log in training data ? (I.e.: does lat,lon belong to this location or somewhere else. Like lat,lon indicate job applicant's location but output location_id is job's location) $\endgroup$ – Shamit Verma Mar 15 at 14:20
  • $\begingroup$ I removed zip, lat, long. Now score for linear reg is 0.9799545999842915 and score for GB is 0.99 $\endgroup$ – Jhon Patric Mar 15 at 14:22
  • $\begingroup$ Okay, difference is not significant. Can you answer previous question (do lat,long belong to location_id) . $\endgroup$ – Shamit Verma Mar 15 at 14:33
0
$\begingroup$

Feature ranking

You still have location_id as a feature when you're trying to predict location_id. So of course that comes out as the "most important," and the other features' importance scores are probably mostly meaningless.

After fixing that, the feature ranking gives you some valuable insight to a problem, and depending on your needs you might drop low-performing variables, etc.

High performance

(I don't think you're actually computing accuracy in either case.) It is extremely surprising to me that a LinearRegression model does so well; most of your variables seem categorical, even the dependent location_id. Unless there's something predictive in the way the ids are actually assigned? How many unique values does location_id have?

Is the location_id the location of the user, or the job (assuming I've gotten the context right)? In either case, if you have many copies of the user/job and happen to split them across the training and test sets, then you may just be leaking information that way: the model learns the mapping user(/job)->location, and happens to be able to apply that to nearly every row in the test set. (That still doesn't make much sense for LinearRegression, but could in the GBM.) This is pretty similar to what @ShamitVerma has said, but doesn't rely on an interpretable mapping, just that the train/test split doesn't properly separate users/jobs.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.