We currently only have holidays as features. How can we respect features that are not just dates but also e.g. time series (like the international gas price, ...)?

According to the task:

Ihre Implementierung muss für die Vorhersage von Benzinpreisen für gegebene Tankstellen und für
die Ausgabe einer Tankstrategie "Plain text"-Dateien in den folgenden beiden Dateiformaten
ausgeben

we should not only output the perfect fueling strategy but also gas price predictions only.

Don't suggest unreasonably low amounts of refilling like 3.3306690738754696e-16 l.

Maybe we can find out something interesting when looking on the first occurrence of a gas station

• Fördermenge
• weltweite Nachfrage
• Wechselkurs des Euro (denn Rohöl wird in Dollar gehandelt)
• von den Mineralölgesellschaften festgelegter Produktpreis

http://www.spiegel.de/auto/aktuell/benzinpreise-aktuell-die-guenstigste-tankstelle-in-ihrer-naehe-a-185826.html

This is extracted from #15 since it needs more work than expected.
Given the near gas stations at the itinerary, the starting time and the tank size from the front-end...

• look up their IDs in our "database" (based on address, name, ...)
• calculate the arrival point at each stations
• predict the price at the arrival time at each station
• compute optimal fueling strategy
• return result to front-end
• optionally: show additional time spent / saved money statistics

"Erstellen Sie Fahrzeugrouten, von denen Sie annehmen, dass Ihr Programm im Vergleich zu denen anderer Teams gute Ergebnisse liefert."

• Show different solutions for different days/times and compare data driven 😄
• Motivate using a risk-assessed real-time strategy

We could use this package https://github.com/DinoTools/python-overpy to run the queries

Refer to the docs for details.

Mismatch between numbering and references of figures.

Compare some gas stations at the Autobahn, with is effected by commuter traffic, economically underdeveloped areas (Brandenburg)...

Section "Zeitreihenanalyse"
"Überlegen Sie sich dazu den möglichen Einfluss von Merkmalen einer Tankstelle auf die Entwicklung ihrer Benzinpreise (z.B. Tankstellen an Autobahnen, Tankstellen die vom Berufsverkehr erfasst werden oder Tankstellen in wenig erschlossenen Regionen)."

Explore in our data what the maximum distance of a gas station to an Autobahn is that we would say it's a "Autobahn gas station"

Prediction on the web server doesn't use multiple cores. It utilises only one core.

Add amount of gas and overall cost at each gas station into the overview on the right side

Is this a comma fault or a very high prediction?

{'end': (48.401030000000006, 9.98764), 'name': ['Aral Tankstelle'], 'start': (52.390530000000005, 13.064540000000001), 'address': ['B 100\n6796 Brehna'], 'overall_price': 42005.76141429667, 'fill_liters': [32.512199237071727], 'prices': [1292.0], 'payment': [42005.76141429667], 'stops': [(51.5537376, 12.1966591)]}

One usually only has to download this data once and this clearly doesn't justify using the LFS bandwith for it.

Add option to set the desired fuel level at the destination, otherwise we'll have a breakdown there

In the moment you can only set the time but not the date

Mail an Ludger

with pyunicorn

Integrate oil and gasoline prices in prediction

Proof of concept notebook: https://github.com/WGierke/informatiCup2018/blob/master/notebooks/1.0-sb-oil_prices.ipynb

Problem: We don't have the future oil price if we predict the gas price at a gas station

for landuse in [industrial, commercial, farmland, farmyard, residential, retail]

rel
  (around:1000, 50.08232, 8.54354)
  ["landuse"="industrial"];
(._;>;);
out body;

http://overpass-turbo.eu/s/tjT

Returns points with coordinates, have to query if gas station is in this polygon

Feature could by noisy

e.g. Given a start and an end point, a tank size and the average consumption, compute the fueling stops for the itinerary showing the fuel stations, how much to fuel and the overall price in the end.

• Show route on map
• Get & show fueling stations on map
• Compute & show perfect fueling strategy

Make some data exploration and write a bit about it

Section "Zeitreihenanalyse"
"Analysieren Sie zunächst die historischen Benzinpreise seit Beginn der Meldepflicht im Jahr 2013"