Chapter 05: RESULTS AND CONCLUSION
5.1 Discussion on the Results Achieved
The project was selected as a motivation for recommending the right medicinal drug as per
the condition of the patients by checking the reviews from the dataset and then proceeded
the project starting with the exploration data analysis phase, followed by the data
preprocessing where we need to make the data easy for further analysis and modelling. In
the data exploration stage, we used the statistical techniques as well as the visualisation
techniques to understand the data and its features.In the same section, we had to find the best
n-grams that could represent the relationship and emotions with the features like rating or
data. The next part of the project was the data preprocessing stage. Here we had to remove
the missing, defective and unwanted data from the set as well as any such condition which
had less than two drugs for recommendation since it won’t be as reliable. In the process of
modelling, to handle the limitations which were in NLP, we decided to use Lightgbm to
overcome it. It is one of the fastest machine learning algorithms which is based on the
decision tree classifier. Alongside, we conducted an emotional analysis or sentiment analysis
using NLTK’s Wordnet and SentiWordnet as well as using a word dictionary. Apart from
this, we also normalised the biassed usefulcount by condition for better efficiency and
reliability. All these steps allowed us to measure the total mean predicted result values for all
the drugs under every condition which would help in recommending the right drug by the
order of its value.
5.2 Application of the Project
Since the pandemic began, many people with problems other than covid have avoided going
to hospitals for a thorough examination in order to avoid coming into contact with the virus.
This has resulted in online treatments and the use of drugs found on the internet.
Furthermore, many doctors and medical students have begun practising medicine and
recommending drugs to patients with limited knowledge and experience, resulting in errors
and mistakes in their judgement and a number of deaths.
To avoid such mistakes, we provide a medicine recommendation system which could help
the doctors or people who want to treat themselves and can be used by them while
prescribing or taking medicines respectively.
5.3 Limitation of the Project
In conclusion, these are the limitations we had during the project.
● Sentiment word dictionaries for sentiment analysis is not a great way since it has low
reliability if the rate of categorised good and bad words are limited. Therefore, we could
have provided a criteria where, if the number of sentiment words was 5 or less, we
could exclude the observations to avoid biassed results.