Özet:
One way to save time and resources in the human recruitment and hiring
process is to post open job positions on the Internet. This allows for a wider reach
and attracts a larger pool of potential candidates. However, the sheer volume of
applications received often creates challenges for hiring managers and companies to
identify the most suitable candidate efficiently.
To address this issue, intelligent tools can be employed, such as deep learning
algorithms and recommender systems. These advanced technologies can expedite the
hiring process and aid in the identification of the right candidate for a particular job.
In this paper, we propose a two-fold algorithmic approach to tackle this problem.
Firstly, we suggest building a Recurrent Neural Network (RNN) classifier for
resume classification. RNNs are a type of neural network specifically designed for
sequence data, making them well-suited for analyzing and extracting relevant
information from resumes. This classifier will help automate the initial screening
process by categorizing resumes based on their suitability for a given job.
Secondly, we propose using cosine similarity to create a resume
recommendation system. By calculating the similarity between the requirements of a
job and the content of a candidate's resume, we can identify the best fit candidates
more efficiently. This recommendation system assists hiring managers in shortlisting
candidates who closely match the job requirements.
To evaluate the performance of our proposed RNN classifier, we assess
several criteria, including accuracy, precision, recall, F-score, and the confusion
matrix. Our experiments demonstrate that the RNN classifier outperforms other
classifiers, such as Gaussian Naive Bayes (GNB), Linear Support Vector Machines
(SVM), and Random Forest (RF), when tested on the same dataset. The same RNN
that we have in our research produced the same performance with BERT on the same
dataset.