Time Series Data¶

In this chapter, we will use a few statistical tools to describe Time Series Data.

Exercise 1: Read Bitcoin Data¶

Load the data in bitcoin.csv into Excel. Create a line plot, so that the time is on the x-axis and the price in EUR is on the y-axis.

Exercise 2: Good and Bad Plots¶

Examine the plots in good_bad_plots2.zip . Discuss which of them are good, which are bad and why.

Exercise 3: Difference¶

Create a new column that contains the difference between a value and the previous one. Create a line plot of the difference.

Exercise 4: Log-Transform¶

Create a new column that contains the logarithm of the price. Create a line plot as well.

Exercise 5: Trend¶

Which of the columns you have (raw, diff, log) comes closest to a linear function? Calculate a linear fit on the data. Determine the slope and the intercept.

Exercise 6: Remove the trend¶

Calculate the residuals: The difference between the data and a linear model.

Exercise 7: Moving average¶

Now smooth the de-trended data using a moving average. Create a new column where you calculate the arithmetic mean over 14 consecutive days. Plot the moving average and the raw data in the same plot.

Exercise 8: Lags¶

Create a new column that is a copy of the de-trended data but shifted down by one rows. Calculate a correlation coefficient between the lag column and the original one.

Exercise 9: Autocorrelation¶

Create more lag columns (about 10). Calculate a correlation coefficient for each of them.

Plot the correlation coefficients for each lag as a bar chart, so that the .

Exercise 10: Resampling¶

Find out how you can resample the data, so that you have one value for each month.

Exercise 11: Seasonality¶

Do you observe any seasonal patterns that occur during specific months?

Exercise 12: Forecasting¶

Discuss the results. How could the statistical tools above help you create a forecast?

How could you determine the accuracy of that forecast?

Getting your own data

If you would like to get the most recent data or examine other cryptocurrencies, do the following:

Visit coingecko public API
Click on the /coins/{id}/market_chart section
Press the Try it out button
Fill in the form and submit it
Download the reponse body

The raw data contains Unix Timestamps in the left column. You might want to convert the dates to a more readable date format.

"""
convert raw historical data from coingecko to pandas
"""
import json
from pprint import pprint
import pandas as pd
from datetime import datetime


j = json.load(open("btc_history.json"))
prices = j["prices"]

def to_timestamp(ts):
    ts = ts/1000
    return datetime.utcfromtimestamp(ts).strftime('%Y-%m-%d %H:%M:%S')

df = pd.DataFrame(prices, columns=["time", "btc"])
df["time"] = df["time"].apply(to_timestamp)
print(df)
df.to_csv("bitcoin.csv", index=False)