Вы находитесь на странице: 1из 16

Medbot: A Drug

Recommendation
System

Capstone Project Report


April 2019 – July 2019
PGP Data Science,Big Data
& Business Analytics

Team Members:
Aishwarya Shetty
Prachi Dhamale
Sankeeta Jha
Sharvari Chavan
Sheryl Saji
1
Table of Contents

1. Introduction

2. Problem Statement

3. Description of the Project

4. Exploratory Data Analysis

5. Models Used

6. Chatbot

7. Deployment

8. References

2
Introduction

Advances in modern medicine have enabled health care to become increasingly precise. This
holds promise for patients, however, the increase in the number of possible treatment options can
make it difficult for physicians to decide on the best for a particular patient. Developing a treatment
recommender system could assist physicians in this task, but due to the high-stakes nature of health
care, it is imperative that medical ethics be integrated into its design. After an overview of medical
ethics, in this project, three key design features are investigated – (1) the selection of machine
learning/deep learning model as a learning algorithm, (2) the exploratory data analysis to check the
most relevant drugs based on conditions,ratings,useful count, and (3) how recommendations are
presented and whether patients should be allowed to opt out of suggestions.

3
Problem Statement
Medbot is a medical recommendation system to understand the human inputs, emotions and
recommend them the relevant drugs based on their health conditions. Through a knowledge graph
allows to correlate various features such as symptoms, signs, patient's habits and the season and thus
display a list of possible recommendation of drugs.

4
Description of the project
● The objective of our project is to build a chatbot to understand human inputs, emotions and
recommend drugs system based on health conditions using the reviews, ratings, useful count
given by users on Drugs.com.
● The data is scrapped from sites like Medscape, Drugs.com and for recommendation we have
used Deep Learning Model, Light GBM and LSTM model.
● The dataset provides patient reviews on specific drugs along with related conditions and a 10-
star patient rating reflecting overall patient satisfaction. The data was obtained by scrapping
pharmaceutical review sites named “Drugs.com”.
● Later a chatbot is built to understand human emotions related to their health condition and
recommend drugs from the recommendation system.

5
Exploratory Data Analysis

Fig 1: The above figure shows top 20 drugs per condition

6
Fig 2: The above figure shows bottom 20 drugs per condition

7
Fig 3: The above figure shows the count of reviews in each year

Fig 4: The above figure shows the count of the highest reviews in each
month of the year
8
Fig 5: The above figure shows the count of highest review on each
day.

Fig 6: The above figure shows the count of highest review in each
season of the year.

9
Fig 7: The above figure shows the pie chart representing the
sentiments of patients.

10
Models Used in the Project
1. We have worked on classifying our reviews into side effects, benefited, comments.
2. This is done on the basis of the rating score given by the user, when the score is 1-4, we have
classified the drug into having side effect, when the score is between 6-10, we classify the
drug to be benefited

3. We have also used Light GBM model which is a fast, distributed, high-performance
gradient boosting framework based on decision tree algorithm, used for ranking,
classification and many other machine learning tasks.

11
4. Finally, we have used LSTM model which is used to predict the next word using a sample.
LSTMs are very powerful in sequence prediction problems because they’re able to store past
information. This is important in our case because the previous review of a drug is crucial in
predicting future drugs for humans.

12
Chatbot
Based on the ranking provided against each medicine a dataset is created
for our chatbot’s reference.
We have used IBM Watson assistant as our platform to create our MedBot.
Document Chatbot requirements:
Intent Utterances Entity Action
Recommend “I need medicine {conditionofuser: Look up the
} condition and
Medicine for
display medicine
“{conditionofuse
r:}”

Our aim with an NLP based chatbot is to identify condition based on user’s
layman content.

Intent creation:

13
Entity creation:

Dialogue flow:

14
Deployment:
We deployed our bot on a WordPress Site:

15
References

 www.drugs.com
 www.medscape.com
 https://towardsdatascience.com/introduction-to-language-models-n-gram-
e323081503d9
 https://papers.nips.cc/paper/6907-lightgbm-a-highly-efficient-gradient-boosting-
decision-tree.pdf
 https://towardsdatascience.com/sentiment-analysis-using-lstm-step-by-step-
50d074f09948
 https://courses.cognitiveclass.ai/courses/course-
v1:IBMDeveloperSkillsNetwork+CB0103EN+2019v2/courseware/407a9f86565c441
89740699636b4fb85/3e97aba8ec0d45df80d35b7d5a57ee00/

16

Вам также может понравиться