IJCRT2303080
IJCRT2303080
IJCRT2303080
org © 2023 IJCRT | Volume 11, Issue 3 March 2023 | ISSN: 2320-
2882
ABSTRACT: As a result of the coronavirus, access to legitimate clinical resources has worsened significantly, including
shortages of specialists and health workers, insufficient equipment, and drug shortages. Due to the emergency of the entire
medical community, many individuals died. Individuals started taking medication themselves without proper consultation due to
unavailability, which made their health condition more serious than usual. A growing number of applications are using machine
learning and innovative work is being done in automation. In this project, a drug recommendation system is presented with the
aim of significantly reducing the burden on specialists. Using patient reviews, we developed a drug recommendation system that
uses various vectorization processes such as Bow, TF-IDF, Word2Vec, and Manual Feature Analysis to predict sentiment, which
can be used to recommend the most appropriate drug for a given disease based on different classification algorithms. AUC,
precision, F1 score and accuracy were used to evaluate the predicted feelings. In emergency situations such as pandemics, floods
or cyclones, a medical referral system can help. In the era of machine learning (ML), recommender systems produce more
accurate, faster, and more reliable clinical predictions at minimal cost. As a result, these systems maintain better performance,
integrity and privacy of patient data in the decision-making process and provide accurate information at all times. Therefore, we
present drug recommendation systems to improve the equity and safety of infectious disease treatment. To reduce side effects,
medications are recommended based on the patient's previous health profile, lifestyle and habits. A system like this could be useful
in recommending safe drugs to patients, especially during medical emergencies.
I. INTRODUCTION
With the exponential increase in the number of coronavirus cases, nations are facing a shortage of doctors, especially in rural
areas where the number of specialists is less compared to urban areas. It takes a doctor approximately 6 to 12 years to obtain the
necessary qualifications. Therefore, the number of doctors cannot be rapidly expanded in a short period of time. The framework
of telemedicine should be strengthened as much as possible in this difficult time [1].
With the exponential development of the web and the web based business industry, item reviews have become a necessary and
integral factor in acquiring items worldwide. Individuals all over the world adapt to analyze reviews and websites first before
deciding to buy a thing. While most past surveys have focused on evaluation expectations and proposals in the field of e-
commerce, the field of medical care or clinical therapies has rarely been addressed.
There has been an increase in the number of individuals concerned about their health and seeking diagnosis online. According
to a 2013 Pew American Research Center survey [5], roughly 60% of adults have searched for health-related topics online, and
about 35% of users have searched for diagnoses of health conditions on the Web. A Medicines Recommendation Framework is
really essential with the goal that it can help specialists and help patients build their knowledge of medicines for specific medical
conditions. A recommender framework is a conventional system that suggests an item to the user depending on its benefit and
IJCRT23030 International Journal of Creative Research Thoughts (IJCRT) a1
www.ijcrt.org © 2023 IJCRT | Volume 11, Issue 3 March 2023 | ISSN: 2320-
necessity. These frameworks use customer surveys to 2882
dissect their sentiment and design recommendations for their exact need. In
a drug recommendation system, medicine is offered under specific conditions dependent on patient reviews using sentiment
analysis and feature engineering. Sentiment analysis is the development of strategies, methods, and tools for discerning and
extracting emotional data, such as opinions and attitudes, from language [7].
On the other hand, Featuring engineering is the process of creating more features from existing ones; improves the
performance of models. This thesis is divided into five segments: Introduction section, which provides a brief overview of the
need for this research, Related papers segment provides a brief overview of previous examinations in this field of study,
Methodology section includes the methods used in this research, Results segment evaluates the results of the applied model using
different metrics, the Discussion section contains the limitations of the framework and finally the conclusion section.
The world is experiencing a shortage of doctors due to the exponential increase in coronavirus cases, especially in rural areas
where there are fewer specialists than in urban areas. A doctor must complete his education between six and twelve years. As a
result, it is not possible to add more doctors in a short time. The infrastructure for telemedicine needs to be upgraded as soon as
possible at this difficult time.
III. OBJECTIVES
The sector is experiencing a shortage of doctors because of the exponential growth in coronavirus instances, mainly in rural
areas where there are fewer specialists than in urban regions. A health practitioner should complete his training between six and
twelve years. As a end result, it is not viable to feature extra doctors in a quick time. The infrastructure for telemedicine desires to
be upgraded as quickly as possible at this tough time.
Witch CM et al. [1] The work in this article focuses on pharmaceutical errors, which are reviewed for the general practitioner,
with an emphasis on terminology, definitions, incidence, risk factors, disclosure and legal implications. A number of variables can
contribute to medication errors, including those related to the drug, the patient, and the health care provider. One or more of the
outcomes that doctors may face after making medication errors include losing the trust of their patients, civil lawsuits, criminal
charges, and medical board discipline. Various approaches have been tried in the prevention of pharmaceutical errors with
varying degrees of success. Physicians' ability to provide safe care to their patients can be improved by learning more about
medication errors. Bartlett JG et al.
[2] In the more than 10 years since the last Community-Acquired Pneumonia (CAP) proposal from the American Thoracic
Society (ATS) / Infectious Diseases Society of America, the guideline development process has changed and new clinical data
(IDSA) have been created. Given the proliferation of information on diagnostic, treatment, and management decisions for the care
of patients with CAP, we intentionally limited the scope of this framework to cover judgments ranging from the medical
diagnosis of pneumonia to the discontinuation of antibiotic therapy and the wearing of chest imaging. T. N. Tekade et al.
[3]This article offers a brief summary of facet mining methods as they are used in the search for new drugs. It is essential for
the pharmaceutical industry to carry out research aimed at detecting adverse drug reactions as quickly as possible. It is a difficult
task to identify important themes from short and noisy reviews. As a solution to this problem, a Probabilistic Aspect Mining
Model (PAMM) is proposed to find aspects and objects related to class labels. Due to the special characteristic of PAMM, it
focuses on discovering features specific to one class rather than simultaneously discovering features for all categories during each
operation. Doulaverakis et al.
V. PROPOSED METHODOLOGY
The above discern determines the device structure of the proposed machine. The system architecture involves following steps:
A. Data Collection and Preprocessing Machine
Learning needs models and lots of data to work. The process of collecting signals that monitor actual physical situations and
converting the obtained results into electronic integer values that can be manipulated by a computer is known as data acquisition.
The processing of primary data includes subsequent procedures. In order to compare the details of individual responses, it is
necessary to combine a huge amount of raw data obtained from field investigations. The method for transforming dirty data into
clean datasets is known as data preprocessing. Real-world information is consistently inaccurate and lacks specific behaviors or
patterns. It is also often inconsistent and incomplete.
Whether we are shopping, buying products online or eating out, reviews are gradually becoming part of our daily routine. We
use reviews to help us make the best decisions. Multiple machine learning techniques were used to construct the recommender
system, which includes Perceptron, Multinomial Naive Bayes, Logistic Regression, Ridge classifier, and Linear SVC
implemented on TF-IDF, Bow, and classifiers such as LGBM, Decision Tree, and Random Forest. Our examination of the
models using five main metrics: f1score, validity, recall, precision, and AUC score shows that linear SVC using TF-IDF
outperforms all other models with 93 percent accuracy. On the other hand, the Word2Vec decision tree algorithm fared the worst,
achieving only 78% accuracy. We integrated the highest expected sentiment values from each strategy LGM on Word2Vec
(91%) Perceptron on Bow (91%) Random Forest on manual features (88%) Linear SVC on TF-IDF (93%) and combined them
according to the standardized number of useful to create referral system. This gave us a total drug score for each condition.
To increase the effectiveness of the recommender system, future work will evaluate different resampling techniques, use
alternative n-gram values, and simplify the
algorithms.
The unborn work involves comparing different slice shapes using different values of n-grams and optimizing algorithms to
improve the performance of the recommender system.
VIII. RESULT
The drug review sample used in this study was obtained from the UCI ML resource. This data consists of six components: the
name of the drug used, the patient's rating, the patient's condition, a valuable number that indicates the number of people who
experienced the benefit of the rating, the date the review was written, and a 10-star patient rating that indicates how the patient is
doing overall satisfactory. According to the user's star rating, each review in this work has been categorized as positive or
negative. Positive reviews are reviews with five or more stars, while negative reviews vary from one to five stars. In Figure 2, we
can see the top medical conditions with the largest number of treatment options. One factor to observe in this figure is the fact that
there are two green bars that indicate criteria that are of little importance.
.
Figure 2: Most Recommended Drugs Per Conditions
Figure 4: Displays The Top Four Medications That Our Algorithm Recommends for The Five Top Medical Issues Including
Acne, Contraception, High Blood Pressure, Anxiety and Depression.
IX. REFERENCES
[1] Wittich CM, Burkle CM, Lanier WL. Medication errors: an overview for clinicians. Mayo Clin Proc. 2014 Aug;89(8):1116-
25.
[2] Bartlett JG, Dowell SF, Mandell LA, File TM Jr, Musher DM, Fine MJ. Practice guidelines for the management of
community-acquired pneumonia in adults. Infectious Diseases Society of America. Clin Infect Dis. 2000 Aug;31(2):347-82.
[3] T. N. Tekade and M. Emmanuel, “Probabilistic aspect mining approach for interpretation and evaluation of drug reviews,”
2016 International Conference on Signal Processing, Communication, Power and Embedded System (SCOPES), Paralakhemundi.
[4] Doulaverakis, C., Nikolaidis, G., Kleontas, A. et al. GalenOWL: Ontology-based drug recommendations discovery. J
Biomed Semant 3, 14 (2012).
[5] Gao, Xiaoyan, Fuli Feng, Heyan Huang, Xian-Ling Mao, Tian Lan, and Zewen Chi. “Food recommendation with graph
convolutional network.” Information Sciences 584 (2022): 170-183.
[6] Chen, Yu-Xiu, Li-Chih Wang, and Pei-Chun Chu. “A medical dataset parameter recommendation system for an autoclave
process and an empirical study.” Procedia Manufacturing 51 (2020): 1046-1053.
[7] Fox, Susannah and Duggan, Maeve. (2012). Implementing a Machine Learning Model to Realize an Effective IOMT-
Assisted Client Nutrition Recommender System. Pew Research Internet Project Report.
[8] J. Ramos et al., “Using tf-idf to determine word relevance in document queries,” in Proceedings of the first instructional
conference on machine learning, vol. 242, pp. 133-142, Piscataway, NJ, 2013.
[9] N. V. Chawla, K. W. Bowyer, L. O. Hall and W. P. Kegelmeyer. SMOTE: Synthetic Minority Over-sampling Technique,
2011, Journal of Artificial Intelligence Research, Volume 16, 2020
[10] Leilei Sun, Chuanren Liu, Chonghui Guo, Hui Xiong, and Yanming Xie. 2016. Data-driven Automatic Treatment
Regimen Development and Recommendation. In Proceedings of the 22nd ACM SIGKDD International Conference on
Knowledge Discovery and Data Mining (KDD ’16). Association for Computing Machinery, New York, NY, USA, 1865-1874.
[11] V. Goel, A. K. Gupta and N. Kumar, “Sentiment Analysis of Multilingual Twitter Data using Natural Language
Processing,” 2018 8th International Conference on Communication Systems and Network Technologies (CSNT), Bhopal, India,
2018.
[12] Shimada K, Takada H, Mitsuyama S, et al. Drug recommendation system for patients with infectious diseases. AMIA
Annu Symp Proc. 2005;2005:1112.