Reference no: EM133296811
Your company makes smart parking meters. The parking meters have a camera, which can be used to take photos of license plates. Your client is a city. It has many plans and possibilities for these parking meters, but for the time being, it issues automatic citations if someone overstays their visit.
The city is primarily interested in compliance with the parking regulations, not citation revenue. For our scenario, the city has agreed to pay for the initial installation of the meters and give you a yearly bonus for any meter with fewer than 50 overstays per year. The bonus in itself is large enough that your company is perfectly content with that bounty being the sole source of revenue related to smart parking meters for the city.
The economics of the smart parking meter is that you make a profit of $3 per citation, and you have to pay the city $20 for every incorrect citation that it issues. For the smart parking meter business to be practical for your company, the value threshold is $100/year (per meter) or getting the bonus from the city, which is the preferred choice. The best current estimate is that there are at least 300 overstays per year for each parking spot, with the maximum profit per meter at $900 if all citations are correctly issued.
Question:
Using the smart parking meter example develop the formula for determining if your ML pipeline is economically viable. Calculate the minimum accuracy of the license plate recognition if the city reduced its hedge to 40 overstays per year. Share your formula, show us the math, and suggest opportunities to capitalize on the reduced hedge.