robertcaliforniadk/danish-car-marketplace-dataset

Danish used car listings — raw dataset

~28,000 car listings scraped from the danish market as of june 2026. this is the raw, messy version — unit suffixes mixed into values, danish decimal separators, empty fields, the whole thing. the point is to have something real to practice data cleaning and analysis on, not a tidy kaggle dataset that does half the work for you.

if you want to jump straight into modeling with this data, check out danish-used-car-price-prediction — that repo walks through loading, cleaning, eda, and training a few models to predict listing prices.

what's in the file

raw_cars_filtered.csv — 28,303 rows, 17 columns.

column	what it is	example
model	full model name (brand + variant + trim)	VW Passat 2,0 TDi 170 Highline Variant DSG BMT 5d
pris	asking price as a string	39.900 kr
pris_int	price as integer (dkk)	39900
modelar	model year	2011
registrering	first registration (month/year)	5/2011
kilometertal	mileage	365.000 km
drivmiddel	fuel type	Benzin / Diesel / El / Plug-in / Hybrid
braendstofforbrug	fuel economy	18,9 km/l
periodisk_afgift	periodic tax	Ejerafgift 4.960 kr/år
ydelse	horsepower	170 hk
acceleration	0–100 km/h	Accel. 8,8 sek
geartype	transmission	Automatisk gear / Manuelt gear
traekvaegt	towing capacity	1.800 kg trækvægt
doere	number of doors	5
scraped_at	when it was pulled	2026-06-20 10:42:37

the data at a glance

• prices from 1,500 kr up to 12,000,000 kr (median ~170k)
• model years 2006 → 2026
• mileage anywhere from 1,000 to 2,000,000 km
• fuel split: 10.7k benzin · 10.4k electric · 4.5k diesel · 2.2k plug-in hybrid · 450 hybrid
• transmissions: ~19k automatic, ~9.4k manual
• mostly 5-door cars

what makes it "dirty"

the values aren't separated from their units — everything comes as a single string. you'll need to parse stuff like:

"365.000 km"     →  365000
"39.900 kr"      →  39900
"170 hk"         →  170
"Accel. 8,8 sek" →  8.8
"18,9 km/l"      →  18.9

also the danish locale uses . as a thousands separator and , as a decimal — so 18,9 means 18.9 and 365.000 means 365000. easy to trip on if you're not paying attention.

getting started

import pandas as pd

df = pd.read_csv("raw_cars_filtered.csv")
df.head()

everything is in danish so you'll want to map column names if that bothers you:

col_map = {
    "pris": "price",
    "pris_int": "price_dkk",
    "modelar": "model_year",
    "registrering": "registered",
    "kilometertal": "mileage_km",
    "drivmiddel": "fuel",
    "braendstofforbrug": "fuel_economy",
    "periodisk_afgift": "tax",
    "ydelse": "hp",
    "acceleration": "accel_0_100",
    "geartype": "transmission",
    "traekvaegt": "towing_capacity",
    "doere": "doors",
}
df = df.rename(columns=col_map)

project ideas

build a price predictor

train a model to estimate listing prices from mileage, year, horsepower, and fuel type. the pris_int column has a clean integer for you to target. tree-based models (random forest, xgboost) work well on this kind of data — r² around 0.89, off by ~25k dkk on average.

electric vs diesel vs gas deep-dive

with ~10k electric and ~10k gasoline listings side by side, you can ask stuff like:

• how does depreciation differ between fuel types?
• what does the electric premium look like at different price brackets?
• do ev owners put on fewer km per year?

depreciation curves

plot price against age, split by brand or fuel type. fit a curve. compare how fast an audi holds value vs a renault, or an electric vs a diesel.

data cleaning pipeline

write a reusable script that takes this raw csv and spits out a clean, analysis-ready version. good practice for structuring a data project properly — config, parsing, validation, output.

dashboard

throw it into a streamlit app. let people filter by brand, fuel, price range, and see what's on the danish market right now.

brand & market analysis

extract brand from the model column. figure out who dominates each price tier, how the market changed over model years, what features correlate with higher prices.

things to watch out for

• a handful of listings have weird prices (12m dkk for what's probably a typo)
• some fields are just empty
• mileage has a few cars over 1.5m km — outliers or data entry errors?
• the scraped/updated/expires timestamps are datetime, but registrering is just month/year as a string

related

• danish-used-car-price-prediction — example project using this data: cleaning, EDA, and modeling car prices