Desert Flow

Finding solutions to the water crisis in the Navajo Region using machine learning and linear regression.

Comment 1

Inspiration

Environmental Impact and Machine Learning where two topics that kept coming to mind when brainstorming ideas.

What it does

We created a model that analyzes historical data given from the Navajo reservoir to determine predictions of outcomes for when reservoir's net storage volume is expected to reach zero if sustainable ideas are not implemented and practiced, based on current inflow, outflow, and overall trends. The model also gives possible solutions to budget water consumption and data trend to restore the level of water at a sustainable capacity in the Navajo Region.

How we built it

Within Google Colab, we used Python along with libraries such as Pandas for reading and processing the CSV files, NumPy for numerical operations, Matplotlib and Seaborn for data visualization, and Scikit-learn for building and training the machine learning model. These tools helped us explore and analyze the reservoir data, generate insights, and build predictive models.

For showcasing the results, we developed a website using HTML, CSS, and JavaScript, where we embedded the visualizations and displayed the output of the machine learning model predictions.

Challenges we ran into

Some of the CSV files were missing information due to gaps in the recording or measurement dates. This required us to clean, preprocess, and handle missing, duplicate, or inconsistent data before analysis. Additionally, we faced challenges in implementing other machine learning models, as their predictions were not reasonable or accurate for our dataset, which limited us to using a linear regression model.

We also struggled to find any relevant open datasets for comparison or further validation. Another major challenge was dealing with inconsistent timelines among the features, which complicated the data merging process and required special handling to align the datasets.

Accomplishments that we're proud of

We’re proud of the solutions our model provided based on the available data. We successfully calculated the optimal water consumption required for reservoir levels to stabilize. By using linear regression, our model predicted the timeline for water depletion and displayed it through a graph, offering a clear and actionable forecast of how long it would take for the water levels to return to normal, along with the rate of reduction. This accomplishment highlights our ability to use data-driven insights to address real-world challenges.

What we learned

Throughout this project, we learned how crucial data preprocessing is, especially when dealing with incomplete or inconsistent datasets. We gained deeper insights into handling missing data, outliers, and creating new features to improve model accuracy. Additionally, we explored different machine learning models, understanding their limitations and why linear regression was the most suitable for our use case. We also strengthened our skills in data visualization and learned how to effectively showcase our results on a website using HTML, CSS, and JavaScript. Overall, this project enhanced our ability to apply machine learning to solve real-world problems.

What's next for DesertFlow

We aim to enhance DesertFlow by creating a more interactive user interface and improving the overall user experience. Some key improvements include:

  1. Developing more interactive data visualizations, allowing users to engage with the graphs, explore trends, and customize views based on their interests.
  2. Implementing real-time updates on the webpage, enabling the model to process live data from the reservoir and provide users with up-to-date predictions and insights.

These enhancements will make the platform more dynamic and user-friendly, allowing for better decision-making and awareness of water conservation efforts.

Share this project:

Updates