Lightbulb Companion

• 

• 

  Items frequently bought together, and inventory checker

• 

  Select available items

• 

  Real-time visualisations

• 

  Predicting profits and sales

• 

  Product demand forecast by item and category

Inspiration

Inspired by NCR's challenge at HackGT, we wanted to create a dashboard that provides small business owners with insights into their sales data. Through this project, we wanted to help management make intelligent, data-driven decisions, and gain a competitive advantage.

What it does

We performed predictive analysis on retail data based on NCR's Silver API endpoints. We then presented these findings in an interactive, intuitive and robust user interface. We were able to:

• deliver insights on items frequently bought together
• use NCR's Inventory API, and our predictive model, to alert owners to dwindling inventory
• use historical trends to predict future sales and profit

How we built it

We used the POS API to get data, and performed analysis of this historical data using pandas, scikit-learn, and other machine learning libraries. Algorithms used include time series analysis and exponential smoothening. Finally, we combined Node.js and Python to create our robust, cross-platform web application. This web app includes integrations with Tableau and Anychart for dynamic, interactive data visualisations.

Challenges we ran into

The NCR Silver API was not always reliable, so we generated historical data based on NCR's API, and stored it locally. In addition, preprocessing huge datasets took additional time and effort, and led to some memory issues. We also had to learn about new machine learning algorithms to apply to our datasets. Also, since our web app used a combination of Node.js and Python, we faced some difficulties in integrating them.

Accomplishments that we're proud of

We made it!!!

What we learned

Big data programming, and use of machine learning algorithms on industry-specific datasets.

What's next for Lightbulb Companion

We want to make this app and its analysis entirely real-time, based on live data from the API. We could also add more predictive models, with greater accuracy, to deliver more insights to business owners, and add predictive inventory management capabilities.

Built With

• any-chart
• css
• exponential-smoothing
• express.js
• html
• html5-up
• javascript
• jquery
• jupyter-notebook
• machine-learning
• ncr-pos-api
• node.js
• numpy
• pandas
• postman
• python
• scikit-learn
• tableau
• time-series-analysis
• unsupervised-learning