The loading page
The home/search page
We are a group of three high school juniors at TJHSST.
Too many times, we've purchased an item we saved money for, only to find that the price of the item had gone down within a week. We wanted to fix this problem by projecting future prices and an ideal buying time.
What it does
Groundhog allows the user to search for a product, stock, or cryptocurrency. It then displays relevant information about the product, including a monthly or weekly graph, projected future lows and highs, and a recommended purchase time. It then allows the user to choose from the sellers analyzed to purchase the product.
How we built it
We used the Adobe Creative Suite to handle branding and mock-ups, then translated these into a full-fledged app using Android Studio, working with java and xml. In order to create the predictive models and recommendations, we used a trial set of past price trends, incorporating a decision tree to maximize the accuracy of our function, or 'concept', simply based around the minimization of the error rate. This function took in an observation vector, x, with preset features, and outputted the response variable, y.
Challenges we ran into
One of the greatest challenges were the issues with Android Studio. it was very difficult to work with this IDE, and we experienced multiple significant delays because of this. Training an algorithm to accurately predict prices also proved to be challenging, and we can't determine its accuracy in the real market since the prices of items, stocks, and cryptos have not changed significantly since when we started the project.
Accomplishments that we're proud of
We're proud of the name, groundhog. Groundhogs are famous for being able to predict when the winter ends, which is similar to how our app predicts when prices will fluctuate. We are also proud of our prediction algorithm and generally smooth and simplistic UI which allow the user to focus on what is important instead of getting lost in complicated details and variables.
What we learned
We learned about how to implement a good machine training algorithm, as well as improvements on building a well-designed android app from scratch.
What's next for Groundhog
We could implement more date ranges for more future-minded users. There's always room for more improvement with the learning function; taking more observations from similar items would very likely mean more accurate output.