Insights and Recommendations on Drug Abuse

Drug abuse is a major health issue in the US. Our model Analyzes various factor and predicts the area-wise potential casualties due to drugs.

Comment

Inspiration : Drug abuse is a prevalent and major health issue in today's world. The drug usage in US has doubled over the last decade and requires immediate action. Many counties in Georgia are affected largely by this epidemic. According to research, 9% of the people in US consume drugs on a regular basis. Hence, we have aimed to obtain a deeper understanding and analysis of this growing issue and improve the health of citizens on US.

What it does : It explores the trends of drug abuse related deaths over last two decades and predicts the potential casualties due to drugs in the future.

We also analysed legal documents describing the usage and regulation of controlled substances to observe how laws impact drug usage in states. We have also analysed various factors affecting drug abuse like poverty, unemployment, demographics, laws and regulations and obtained insightful results to recommend and help the society's well being

How I built it: We followed various steps to achieve our model completion:

  • Defining the problem
  • Data gathering
  • Unstructured data with MongoDb
  • Data processing/cleaning
  • Developing Hypothesis
  • Data exploration using Tableau
  • ML modelling using NLTK, fb-prophet
  • Deep learning modelling - LSTM and keras
  • Model evaluation and improvement
  • Insights and recommendations based on outputs and visualisation
  • Hosted on Google Cloud

Challenges I ran into. :

  1. Finding the right data which would solve our problem
  2. Choosing the variables and parameters for time series predictions
  3. Choosing the right model suiting our needs

Accomplishments that I'm proud of:

  1. We found similarities between legal documents using unstructured data which gave good results and insights
  2. Successfully able to predict area wise deaths by drugs with an accuracy 85%

What I learned:

  1. First time, we have executed deep learning and we learnt a lot through this process.
  2. Leveraged google cloud to host tableau server
  3. Learnt to plot interactive charts using tableau

What's next for Insights and Recommendations on Drug Abuse:

  1. Increase in rehabs in particular counties reduces deaths
  2. Age group 20-35 are prone to drugs
  3. Poverty has high correlation with deaths by drugs
  4. Provided various recommendations based on insights obtained to Police, Parents and Government

Built With

Share this project:

Updates