The Challenge :

The coronavirus has emptied out cities and fewer people are driving cars or boarding planes decimating demand for energy around the globe. Fuel burning business like airlines and factories are idled. But many of the world's major oil producers have been pumping more than ever, leading to a crash in oil prices. Price movement in one asset shows some correlation with price movement in other assets, for example, if the price of oil rises, then it will cost more to make plastic, and a plastics company will then pass on some or all of this cost to the consumer, which raises prices and thus creates inflation. Other industries like the Transportation, Supply chain, Logistics and Manufacturing sectors also get affected. COVID-19 has severe negative impacts on human health and the world economy.

What inspired The Butterfly effect?

The Oil bust is reshaping Global Markets, even though consumers like low prices, but in today's environment, where so many people are quarantined, the people cannot benefit from low oil prices, so there's no winner in this current situation. Market turmoil is about more than just the virus. Without clear picture on what is happening on and to the earth, bad business decisions can cause billions of dollars for companies and further destroy our planet.

Our Approach 

Our Analytics platform tool integrates various datasets between Jan 2020 and May 2020. For our initial prototype we decided to focus on 3 areas:

• Prediction of Covid cases in the next week in 10 cities of the United States to help state and local leaders hopefully disrupt the disease cycle.
• Weekly heat maps of NO2 in big cities was converted to an estimate using Vectorisation  from NASA satellite image data 
• How this impacts Retail gasoline prices

We built a single layer 64 neurons LSTM which was implemented to find the correlations between trends and gas prices and NO2 emission in big cities, before and after the peak in the number of cases. We used Keras Tensor flow and Python (sklearn, Keras, Pandas, Numpy, Matplotlib).

The input is the NO2 weekly data and the output is gas prices. 

Analysis

Logically, the behaviors captured by the datasets in our solution are:

• The algorithm predicts a rise in Retail gas prices in big cities where corona cases have already passed their peak of Confirmed COVID cases. In Denver, New York, Miami and Boston there is still no change in retail gas prices when compared to cities where it is still on rise. This shows us that there are other factors affecting the supply and demand of Gasoline in those respective counties.
• The prediction represents direct correlation between NO2 emission and usage of vehicles or other transportation systems. This in turn correlates to voluntary social distancing. The predictions reflect the decrease in social distancing in cites where the peak has passed (in cites of confirmed Covid cases).
• This can be a probable inverse correlation between No2 and gas prices which can be used to predict gas prices in the following weeks in the cities  where the COVID cases are likely to decrease.
• We have plotted multiple charts showing correlations and are available with the code file below.

https://drive.google.com/file/d/1Rfh3FL4yXOQT8fiu7Sl01syVx5ECMNm8/view?usp=sharing

• NO2 satellite data - https://so2.gsfc.nasa.gov/no2/no2_index.html
• Covid19 data - https://github.com/CSSEGISandData/COVID-19#covid-19-data-repository-by-the-center-for-systems-science-and-engineering-csse-at-johns-hopkins-university
• Gas prices - https://www.eia.gov/dnav/pet/pet_pri_gnd_dcus_nus_w.htm