Personality Classification for Recruitment�(Through Data Analysis and Data Classification)

Guided by:

Prof. Mukul Aggarwal

Submitted by:

Shuchita Jain (1602913102)

Shweta Kapoor (1602913103)

Final Project Presentation Group No. LG-7

DEPARTMENT OF INFORMATION TECHNOLOGY

KIET GROUP OF INSTITUTIONS, GHAZIABAD, UTTAR PRADESH

(AFFILIATED TO DR. A.P.J. ABDUL KALAM TECHNICAL UNIVERSITY, LUCKNOW, UTTAR PRADESH, INDIA)

Content

​

• Abstract
• Objective
• Research Paper
• Planning of Work
• Literature Review
• Attribute Selection

​

• Proposed Work
• Implementation
• Results
• Conclusion
• References

Abstract

A personality analysis system is a system in which the data about a person’s traits are collected using some means such as their social accounts, write ups, blogs or direct questions about their personality. The key traits are analyzed and identified using some data mining techniques. This model or system collects data and identify traits of a candidate that can be useful for the company recruiters to select people who are fit for a particular job description on the basis of skills, experience and their overall personality. This is done by identifying essential traits present in an individual by analyzing the personality of a person on the basis of the data collected from surveying. The system analyses the personality of a candidate through some machine learning classification techniques to classify people on the basis of user's characteristic data. This classification can now be used for various purposes but here we use it for the recruitment process done by the HR. It would help recruiters to eliminate incapable candidates and selecting appropriate ones based on their personalities at the initial stages hence reducing the time efforts and cost to a considerable amount.

3

Objective

The model is designed

• To save time and money consumed in organizing recruitment procedure for a large number of candidates.
• To provide a medium for the people who wish to identify which job profile matches as per their traits.
• To classify the candidates in accordance with their personality.
• To provide an efficient medium for recruiters by reducing their workload upto some extent.

4

5

• DISC [14] stands for dominance, influence, steadiness, and conscientiousness.
• It was given by Walter Clark in 1940, which contains 28 questions.

​

​

​

​

​

​

​

DISC

MODELS STUDIED

​

​

​

​

​

​

​

​

• The Values and Motives inventory[19] measures extrinsic and intrinsic values.
• It summarizes all the possible motivating factors and demotivating factors to the work of a person.

​

​

​

​

​

Figure 1. DISC Model [14]

Values and Motives Inventory

​

Figure 2. Values & Motives Inventory [19]

6

6

True Colors

​

• True Colors [21] was constructed by Don Lowry in 1978.
• This test evaluates four basic learning styles: action-oriented, independent thinkers, people-oriented and pragmatic planners.

​

​

​

​

​

Figure 3. True Colors Model [21]

HEXACO

• The HEXACO model [16] assesses the six major personality dimensions.
• HEXACO: Honesty-Humility, Emotionality, Extraversion, Agreeableness, Conscientiousness,, Openness to Experience.

Figure 4. HEXACO Model [16]

7

Eysenck Personality Inventory

​

• The Eysenck Personality Inventory[18] assesses personality only on 2 independent dimensions:
• stability versus neuroticism
• introversion versus extroversion
• It generates 3 scores, the 'N' score, an 'E' score and a 'lie' score

Figure 5. Eysenck Personality Inventory [18]

8

Attribute Selection�

In Attribute Selection, we grouped some relevant personality traits (obtained from comparative analysis) into a single characteristic, which gave us the major seven traits that we have used in our model. �Those 7 traits will be shown in the next slide.

10

Figure 6. Traits Identification

Proposed Work

Our proposed model consists of seven major personality traits (listed in the top row, in next slide) that we are going to look for in a candidate for a particular job. Each column on this table elaborates the corresponding personality trait.

12

Table 1. Proposed Model

IMPLEMENTATION

13

1. Collection of Data

• To collect data, a questionnaire was prepared that could give insights on behavioral aspects of an individual.
• We prepared a set of four questions for each of our 7 traits making a total of 28 questions.
• The nature of the questions and answers was kept in such a manner that there was no right or wrong answer.
• This questionnaire can be answered through a Google form. We collected data from more than 100 people and applied our analysis on it.

Questionnaire

Link (Google Form): https://forms.gle/yEsS3r11VgbkeM188

Dataset

Link (Data collected through Surveying)-https://docs.google.com/spreadsheets/d/1f347PNpt2IvUo2P1T2C6Uh3fWObBnNFjJFABb6OUN5U/edit?usp=sharing

2. Pre-processing of Data

We have done Pre-processing of data in three phases:-

2.1 Job Profiles

• The first task was to design some sample job profiles for which the candidate is to be evaluated.
• These job profiles were designed by assigning a score out of 20 in each trait.
• It depicts the ideal score that the candidate should have to be perfect for the job.
• We have designed 8 such job profiles for this analysis.
• A recruiter can design their own job profile based on their requirement.

Figure. Profiles

2.2 Score Calculation

• The next task was to calculate the score of each individual in the 7 traits.
• This was done through addition of candidate’s form response of each question.
• Each question has a weightage of 5 thus each trait has a weightage of 20 (each trait has 4 questions).

Score calculation of a candidate

This table shows the form responses of particular candidate:-

This table shows the score calculation of a particular candidate based on her responses:-

22

Figure. Classification Based on Scores

This figure shows the score of each individual in the 7 traits, with the profile as per their scores.

2.3 Distance Calculation

Figure. Distance Matrix

The Euclidean distance with each profile was calculated and the profile which has least distance for a particular candidate was assigned to the respective individual.

3. Classification

• We have applied the following techniques of classification:

​

• KNN
• SGD
• Random Forest
• SVM
• Neural Network

​

• These techniques were applied with a 10 folds cross validation.
• For all these techniques, we calculate six parameters- AUC, ROC, CA, F1 Score, Precision and Recall.
• Based on these six parameters we compare the above five techniques or models.

​

Classification Techniques used:

This figure shows the work flow (classification) on Orange tool, using five classification techniques.

Figure. Comparison on Orange Tool

Results

General Analysis

​

Comparison of Techniques

Each of the model was compared on the basis of AUC, ROC, CA, F1 score, Precision, Recall.

Table, shows the analysis of different classification models based on above six parameters.

• kNN
• Random Forest
• Neural Network
• SVM
• SGD

ROC Analysis Curve

The ROC analysis graph-

​

Target class is- Technician Costs: FN = 500, FP = 500 Target probability is-50.0 %

​

(ROC Curves summarize the trade-off between the true positive rate and false positive rate for a predictive model using different probability thresholds.)

• kNN
• Random Forest
• Neural Network
• SVM
• SGD

Lift Curve

The Lift Curve –

​

Target class is- Technician

​

​

​

​

(A lift is measure of the performance of a targeting model at predicting or classifying cases as having an enhanced response, measured against a random choice targeting model.)

Conclusion

• This model classifies applicants on the basis of their overall personality and experience in more precise and efficient manner.
• It maps the traits of an individual with the needs of the recruiter for a particular job description by identifying essential traits present in an individual and analyzing the personality of a person on the basis of the data collected from surveying.
• For this purpose, it considers seven attributes along with the specifications and fundamentals of each attribute that should be required in any candidate for being recruited by the recruiters.
• The model classifies a great extent of user characteristics and behaviors, and able to provide the best suitable job profile that matches with user’s personality.
• In this way, this model fulfills the need of recruiters for hiring the best candidates for a particular job profile. This classification can be used for recruitment process done by the HR.

33

References

[1] Alvaro Ortigosa, Rosa M. Carro, José Ignacio Quiroga. Predicting user personality by mining social interactions in Facebook Computer Science Department, Universidad Autónoma de Madrid, Madrid, Spain Journal of Computer and System Sciences 80 (2014) pp.57–71

[2] Ana Carolina E.S. Lima, Leandro Nunes de Castro. A multi-label, semi-supervised classification approach applied to personality prediction in social media Natural Computing Laboratory, Mackenzie Presbyterian University, São Paulo, Brazil Neural Networks 58 (2014) pp 122–130

[3] Stefan Strohmeier, Franca Piazza. Domain driven data mining in human resource management: Chair of Management Information Systems, Saarland University, 66123 Saarbruecken, Germany Expert Systems with Applications 40 (2013) pp 2410–2420

[4] Gabriel Yakub N.N. Adi, Michael Harley Tandio, Veronica Ong, DerwinSuhartonoOptimization for Automatic Personality Recognition on Twitter in Bahasa Indonesia Computer Science Department, School of Computer Science, Bina Nusantara University, Jakarta, Indonesia 11480 Procedia Computer Science 135 (2018) pp 473–480 1877-

[5] Alfred Kobsa, Generic user modeling systems, in: The Adaptive Web: Methods and Strategies of Web Personalization, in: Lecture Notes in Comput. Sci.,vol. 4321, Springer, Berlin, Heidelberg,( 2007), pp. 136–154.

[6] R. K. Dwivedi, M. Aggarwal, S. K. Keshari, A. Kumar, S. Bhattacharyya, A. Hassanien, D. Gupta, A. Khanna, I. Pan, "Sentiment Analysis and Feature Extraction Using Rule-Based Model (RBM)", International Conference on Innovative Computing and Communications, vol. 56, (2019).

[7] Mukul Aggarwal, Amod kumar Tiwari, Deepak Gupta “Critical Analysis of Information Retrieval Models”, Journal of Advanced Research in Dynamical and Control Systems (JARDCS), Vol 10, Issue 14, ( 2018), pp 1591-1597.

​

​

​

34

References (Cont’d)

[8] Blog: Introduction to Psychology Module 12.

[9] Wikipedia: 1879, Wilhelm Wundt

[10] Wikipedia: Freud's psychoanalytic theories

[11] Wikipedia: Carl Jung

[12] Wikipedia: Woodworth Personal Data Sheet

[13] Blog: Why Hiring For Personality is as essential as Experience

[14] Walter Clark 1940, DiscProfile.com

[15] Raymond Catell, OpenPsychometrics.org

[16] Kibeom Lee, Ph.D., & Michael Ashton, Ph.D., Hexaco.org

[17] Katharine Cook Briggs and Isabel Briggs Myers 1940, Myersbriggs.org

[18] Eysenck Personality Inventory Iluguru.ee

[19] Website: PsyTech.com

[20] Max Kostick, MeritSolutions.com

[21] Don Lowry My-personality-test

​

35

Thank You