The setup to visualize the confusion matrix.
Proof that the accuracy score surpassing the initial 9% accuracy by more than 81% after Kareem successfully tweaked the hyperparameters.
Snippet of relevant data that is utilized in this submission.
Raw data that shows the initial implementation of the confusion matrix.
Further insights from a detailed confusion matrix.
One example of data analytics for the dataset.
We aim to create an inclusive environment for those in the professional industry looking to stay connected, especially now more than ever due to the COVID pandemic keeping everyone at home. By creating this app, we want everyone to stay connected and feel welcome because in any business it is essential to establish an inclusive environment where all employees feel comfortable allowing them to flourish professionally.
What it does
The Lodge connects users across the Atlanta Metro area with a machine learning-powered service. This is a free-to-use platform and interested individuals can easily signup through the app and further their professional careers. The Lodge aims to improve professional branding by keeping like-minded peers connected through the pandemic and for future endeavors.
How we built it
At the beginning, we have a GoogleColab notebook that creates a .pynb for implementing into the Microsoft Azure Machine Learning platform. Then, the code is designed to operate more efficiently in the Microsoft Azure Machine Learning platform because we have access to the compute optimize CPU that allows the code to process through 200 iterations of validating the data faster. We learned that higher optimization speeds will enable the code to run more smoothly for the end-user.
We intend for the user interface (UI) of the app to function as a front-end source of information that the app user can use for finding the best fit for a connection. We want the machine learning results to influence the outcome of a match for the best fit; more specifically the connection that will bring a diverse group of users together. The dataset is refined and prepared to output an accuracy score above 95%. The code has the ability to be ran in a loop with the results being the same. In the future, we plan to incorporate the confidence interval into our findings too.
Challenges we ran into
First, we needed to identify a dataset that has fields that will be the most practical to leverage in a machine learning model, such as the Multi-Layer Perceptron Classifier. Next, the data within the dataset needed to be formatted into integers that corresponded to the original instances of the dataset. Third, the dataset was not reading properly into Microsoft Azure Machine Learning due to some runtime issues with the kernel. Then, we discovered that we need to store the dataset onto a secure database and server, such as a Microsoft Azure SQL Database and Microsoft Azure Virtual Server, respectively. Lastly, due to time restrictions, we could not integrate our Figma design into an interactive app that uses Azure Security Monitoring. Through extensive collaboration and research, we overcame these obstacles, so that we can provide a proof-of-concept and a minimal viable product (MVP).
Accomplishments that we're proud of
Initially, we only achieved an accuracy score of about 9%, but after continuously refining our approach and designing more effective strategies, we achieved a 97.3% for our accuracy.
What we learned
We learned that teamwork allows for faster and more efficient results. We communicated effectively and prioritized the important tasks. We learned how to analyze important pieces of information from official documentation to better understand the code and how to implement the dataset onto Microsoft Azure from GoogleColab. Lastly, we also learned to implement accuracy enhancing strategies to provide the best possible service.
What's next for The Lodge
We want to securely store a database on the cloud for the purpose of enhancing user experience and expanding our capabilities within the app. We want to be able to create more fields that our Multi-Layer Perceptron Classifier can interpret for the purpose of creating a more refined method of connecting individuals. We also want to expand the app's services to more regions and create larger networks capable of connecting more people with common interests.