jackiekirschner / interview_attendance

Predicting Interview Attendance

Aim

A candidate not showing up to a scheduled interview wastes HR time and resources. In this case study, we would like to predict whether or not a candidate will attend a job interview based on 20 categorical features. In this case, we are interested in precision because our aim is to reduce false positives.

Data Source:

Kaggle: Interview Attendance Problem

	Date of Interview	Client name	Industry	Location	Position to be closed	Nature of Skillset	Interview Type	Name(Cand ID)	Gender	Candidate Current Location	Candidate Job Location	Interview Venue	Candidate Native location	Have you obtained the necessary permission to start at the required time	Hope there will be no unscheduled meetings	Can I Call you three hours before the interview and follow up on your attendance for the interview	Can I have an alternative number/ desk number. I assure you that I will not trouble you too much	Have you taken a printout of your updated resume. Have you read the JD and understood the same	Are you clear with the venue details and the landmark.	Has the call letter been shared	Expected Attendance	Observed Attendance	Marital Status	Unnamed: 23	Unnamed: 24	Unnamed: 25	Unnamed: 26	Unnamed: 27
0	13.02.2015	Hospira	Pharmaceuticals	Chennai	Production- Sterile	Routine	Scheduled Walkin	Candidate 1	Male	Chennai	Hosur	Hosur	Hosur	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	No	Single	nan	nan	nan	nan	nan
1	13.02.2015	Hospira	Pharmaceuticals	Chennai	Production- Sterile	Routine	Scheduled Walkin	Candidate 2	Male	Chennai	Bangalore	Hosur	Trichy	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	No	Single	nan	nan	nan	nan	nan
2	13.02.2015	Hospira	Pharmaceuticals	Chennai	Production- Sterile	Routine	Scheduled Walkin	Candidate 3	Male	Chennai	Chennai	Hosur	Chennai	nan	Na	nan	nan	nan	nan	nan	Uncertain	No	Single	nan	nan	nan	nan	nan
3	13.02.2015	Hospira	Pharmaceuticals	Chennai	Production- Sterile	Routine	Scheduled Walkin	Candidate 4	Male	Chennai	Chennai	Hosur	Chennai	Yes	Yes	No	Yes	No	Yes	Yes	Uncertain	No	Single	nan	nan	nan	nan	nan
4	13.02.2015	Hospira	Pharmaceuticals	Chennai	Production- Sterile	Routine	Scheduled Walkin	Candidate 5	Male	Chennai	Bangalore	Hosur	Chennai	Yes	Yes	Yes	No	Yes	Yes	Yes	Uncertain	No	Married	nan	nan	nan	nan	nan

I. Data Cleaning

1. Removed irrelevant columns:

• Name of candidate
• Date of interview

2. Renamed column names
3. Imputed NaN with 0

• e.g., assumed if "Are you clear with venue details?" == NaN --> False

4. Cleaned string values

• e.g., "Walkin" == "Walk in"

5. Dummy coded variables

	permission_start_on_time	hope_no_unsch_meets	call_three_hrs_before	alt_num_given	come_prepared	can_find_interview_loc	call_letter_shared	expected_attendance	observed_attendance	client_name_ANZ	client_name_Aon Hewitt	client_name_Aon hewitt Gurgaon	client_name_Astrazeneca	client_name_Barclays	client_name_Flextronics	client_name_Hewitt	client_name_Hospira	client_name_Pfizer	client_name_Prodapt	client_name_Standard Chartered Bank	client_name_Standard Chartered Bank Chennai	client_name_UST	client_name_Williams Lea	client_name_Woori Bank	industry_BFSI	industry_Electronics	industry_IT	industry_Pharmaceuticals	industry_Telecom	location_Bangalore	location_Chennai	location_Cochin	location_Delhi	location_Gurgaon	location_Hyderabad	location_Noida	position_to_be_closed_AML	position_to_be_closed_Dot Net	position_to_be_closed_Niche	position_to_be_closed_Production- Sterile	position_to_be_closed_Routine	position_to_be_closed_Selenium testing	position_to_be_closed_Trade Finance	interview_type_Scheduled	interview_type_Scheduled_Walkin	interview_type_Walkin	gender_Female	gender_Male	candidate_current_location_Bangalore	candidate_current_location_Chennai	candidate_current_location_Cochin	candidate_current_location_Delhi	candidate_current_location_Gurgaon	candidate_current_location_Hyderabad	candidate_current_location_Noida	candidate_job_location_Bangalore	candidate_job_location_Chennai	candidate_job_location_Cochin	candidate_job_location_Gurgaon	candidate_job_location_Hosur	candidate_job_location_Noida	candidate_job_location_Visakapatinam	interview_venue_Bangalore	interview_venue_Chennai	interview_venue_Cochin	interview_venue_Gurgaon	interview_venue_Hosur	interview_venue_Hyderabad	interview_venue_Noida	candidate_native_location_Agra	candidate_native_location_Ahmedabad	candidate_native_location_Allahabad	candidate_native_location_Ambur	candidate_native_location_Anantapur	candidate_native_location_Baddi	candidate_native_location_Bangalore	candidate_native_location_Belgaum	candidate_native_location_Bhubaneshwar	candidate_native_location_Chandigarh	candidate_native_location_Chennai	candidate_native_location_Chitoor	candidate_native_location_Cochin	candidate_native_location_Coimbatore	candidate_native_location_Cuttack	candidate_native_location_Delhi	candidate_native_location_Faizabad	candidate_native_location_Ghaziabad	candidate_native_location_Gurgaon	candidate_native_location_Hissar	candidate_native_location_Hosur	candidate_native_location_Hyderabad	candidate_native_location_Kanpur	candidate_native_location_Kolkata	candidate_native_location_Kurnool	candidate_native_location_Lucknow	candidate_native_location_Mumbai	candidate_native_location_Mysore	candidate_native_location_Nagercoil	candidate_native_location_Noida	candidate_native_location_Panjim	candidate_native_location_Patna	candidate_native_location_Pondicherry	candidate_native_location_Pune	candidate_native_location_Salem	candidate_native_location_Tanjore	candidate_native_location_Tirupati	candidate_native_location_TricVellorehy	candidate_native_location_Trichy	candidate_native_location_Trivandrum	candidate_native_location_Tuticorin	candidate_native_location_Vijayawada	candidate_native_location_Visakapatinam	candidate_native_location_Warangal	marital_status_Married	marital_status_Single
0	True	True	True	True	True	True	True	True	False	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	1	0	0	1	0	0	0	0	0	0	0	0	1	0	0	0	0	1	0	0	1	0	1	0	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	1
1	True	True	True	True	True	True	True	True	False	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	1	0	0	1	0	0	0	0	0	0	0	0	1	0	0	0	0	1	0	0	1	0	1	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	1
2	False	False	False	False	False	False	False	False	False	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	1	0	0	1	0	0	0	0	0	0	0	0	1	0	0	0	0	1	0	0	1	0	1	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	1
3	True	True	False	True	False	True	True	False	False	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	1	0	0	1	0	0	0	0	0	0	0	0	1	0	0	0	0	1	0	0	1	0	1	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	1
4	True	True	True	False	True	True	True	False	False	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	1	0	0	1	0	0	0	0	0	0	0	0	1	0	0	0	0	1	0	0	1	0	1	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0

II. Build Models

We built the following models using sklearn:

1. DecisionTreeClassifier
2. RandomForestClassifier
3. BaggingClassifier
4. GradientBoostingClassifier
5. AdaBoostClassifier
6. "MegaUltraClassifier"

Cross-validation

def k_folds_CV(model, n_splits):
    scores = []
    kf = KFold(n_splits)
    for train_index, test_index in kf.split(X):
        Xk_train, Xk_test = X[train_index], X[test_index]
        yk_train, yk_test = y[train_index], y[test_index]
        model.fit(Xk_train, yk_train)
        preds = model.predict(Xk_test)
        score = precision_score(preds, yk_test)
        scores.append(score)
    cv_score = np.mean(scores)
    return cv_score

III. Model Selection

Calculated precision scores for each model:

Model
Decision Tree	.70
Random Forest	.74
Bagging	.73
kNN	.59
GradientBoost	.75
AdaBoost	.77

Grid Search

We selected the AdaBoost model based on highest precision scores, then conducted grid search to find the best parameter settings. We built a new model with those parameters. CV Best Ada Precision: 0.83

Feature Importance

1. Have you obtained the necessary permission to start on time? 0.36
2. Interview Type Scheduled 0.14
3. Position to be closed Niche: 0.08 (Niche refers to rare skill sets while routine refers to more common skill sets)
4. Position to be closed Production- Sterile .02
5. Client Name ANZ: 0.04
6. Candidate Native Location Ahmedabad 0.02
7. Candidate Native Location Vijayawada 0.02

Conclusion

Our cross-validated AdaBoost model with optimized parameters predicts with 83% precision.

Future Steps

1. Find optimal parameters across all models and compare across models
2. Look at partial dependence plots
3. Add logistic regression