{ "cells": [ { "cell_type": "code", "execution_count": 1, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "Päivitetty 2024-03-30 / Aki Taanila\n" ] } ], "source": [ "from datetime import datetime\n", "print(f'Päivitetty {datetime.now().date()} / Aki Taanila')" ] }, { "cell_type": "code", "execution_count": 2, "metadata": {}, "outputs": [], "source": [ "import pandas as pd" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "# Puuttuvat tai päällekkäiset aikaleimat ja havainnot\n", "\n", "Aikasarjojen analysoinnissa ja aikasarjaennustamisessa puuttuvat ja päällekkäiset havainnot aiheuttavat ongelmia.\n", "\n", "## Puuttuvat havainnot\n", "\n", "Tarkastelen esimerkkinä kuukauden ensimmäisten päivien muodostamaa aikasarjaa, josta puuttuu välistä kahden kuukauden (huhtikuu ja toukokuu) tiedot." ] }, { "cell_type": "code", "execution_count": 3, "metadata": {}, "outputs": [ { "data": { "text/html": [ "
\n", "\n", "\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
datevalue
02019-01-012
12019-02-013
22019-03-014
32019-06-017
42019-07-018
52019-08-019
\n", "
" ], "text/plain": [ " date value\n", "0 2019-01-01 2\n", "1 2019-02-01 3\n", "2 2019-03-01 4\n", "3 2019-06-01 7\n", "4 2019-07-01 8\n", "5 2019-08-01 9" ] }, "execution_count": 3, "metadata": {}, "output_type": "execute_result" } ], "source": [ "# Avaan datan:\n", "df1 = pd.read_excel('https://taanila.fi/aika1.xlsx')\n", "df1" ] }, { "cell_type": "code", "execution_count": 4, "metadata": {}, "outputs": [ { "data": { "text/html": [ "
\n", "\n", "\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
value
date
2019-01-012
2019-02-013
2019-03-014
2019-06-017
2019-07-018
2019-08-019
\n", "
" ], "text/plain": [ " value\n", "date \n", "2019-01-01 2\n", "2019-02-01 3\n", "2019-03-01 4\n", "2019-06-01 7\n", "2019-07-01 8\n", "2019-08-01 9" ] }, "execution_count": 4, "metadata": {}, "output_type": "execute_result" } ], "source": [ "# Siirrän aikaleimat indeksiin:\n", "df1.index = pd.to_datetime(df1['date'])\n", "df1 = df1.drop('date', axis=1)\n", "df1" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "**resample()** lisää puuttuvat kuukaudet indeksiin. Aikafrekvenssi **'MS'** tarkoittaa kuukauden ensimmäisiä päivä (**M**onth**S**tart). Eri tilanteisiin sopivia aikafrekvenssejä löydät osoitteesta https://pandas.pydata.org/pandas-docs/stable/user_guide/timeseries.html#offset-aliases.\n", "\n", "**ffill()** täydentää puuttuvat arvot niitä edeltävällä arvolla." ] }, { "cell_type": "code", "execution_count": 5, "metadata": {}, "outputs": [ { "data": { "text/html": [ "
\n", "\n", "\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
value
date
2019-01-012
2019-02-013
2019-03-014
2019-04-014
2019-05-014
2019-06-017
2019-07-018
2019-08-019
\n", "
" ], "text/plain": [ " value\n", "date \n", "2019-01-01 2\n", "2019-02-01 3\n", "2019-03-01 4\n", "2019-04-01 4\n", "2019-05-01 4\n", "2019-06-01 7\n", "2019-07-01 8\n", "2019-08-01 9" ] }, "execution_count": 5, "metadata": {}, "output_type": "execute_result" } ], "source": [ "df2 = df1.resample('MS').ffill()\n", "df2" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "**bfill()** täydentää puuttuvat arvot niitä seuraavalla arvolla." ] }, { "cell_type": "code", "execution_count": 6, "metadata": {}, "outputs": [ { "data": { "text/html": [ "
\n", "\n", "\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
value
date
2019-01-012
2019-02-013
2019-03-014
2019-04-017
2019-05-017
2019-06-017
2019-07-018
2019-08-019
\n", "
" ], "text/plain": [ " value\n", "date \n", "2019-01-01 2\n", "2019-02-01 3\n", "2019-03-01 4\n", "2019-04-01 7\n", "2019-05-01 7\n", "2019-06-01 7\n", "2019-07-01 8\n", "2019-08-01 9" ] }, "execution_count": 6, "metadata": {}, "output_type": "execute_result" } ], "source": [ "df3 = df1.resample('MS').bfill()\n", "df3" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "**interpolate()** interpoloi puuttuvat arvot niitä edeltävän ja seuraavan arvon perusteella." ] }, { "cell_type": "code", "execution_count": 7, "metadata": {}, "outputs": [ { "data": { "text/html": [ "
\n", "\n", "\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
value
date
2019-01-012.0
2019-02-013.0
2019-03-014.0
2019-04-015.0
2019-05-016.0
2019-06-017.0
2019-07-018.0
2019-08-019.0
\n", "
" ], "text/plain": [ " value\n", "date \n", "2019-01-01 2.0\n", "2019-02-01 3.0\n", "2019-03-01 4.0\n", "2019-04-01 5.0\n", "2019-05-01 6.0\n", "2019-06-01 7.0\n", "2019-07-01 8.0\n", "2019-08-01 9.0" ] }, "execution_count": 7, "metadata": {}, "output_type": "execute_result" } ], "source": [ "df4 = df1.resample('MS').interpolate()\n", "df4" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## Päällekkäiset havainnot\n", "\n", "Tarkastelen seuraavaksi esimerkkiä päällekkäisistä havainnoista. Seuraavassa esimerkissä on puuttuvien kuukausien (huhtikuu ja toukokuu) lisäksi päällekkäinen havainto helmikuun kohdalla." ] }, { "cell_type": "code", "execution_count": 8, "metadata": {}, "outputs": [ { "data": { "text/html": [ "
\n", "\n", "\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
datevalue
date
2019-01-012019-01-012
2019-02-012019-02-013
2019-02-012019-02-015
2019-03-012019-03-014
2019-06-012019-06-017
2019-07-012019-07-018
2019-08-012019-08-019
\n", "
" ], "text/plain": [ " date value\n", "date \n", "2019-01-01 2019-01-01 2\n", "2019-02-01 2019-02-01 3\n", "2019-02-01 2019-02-01 5\n", "2019-03-01 2019-03-01 4\n", "2019-06-01 2019-06-01 7\n", "2019-07-01 2019-07-01 8\n", "2019-08-01 2019-08-01 9" ] }, "execution_count": 8, "metadata": {}, "output_type": "execute_result" } ], "source": [ "df5 = pd.read_excel('https://taanila.fi/aika2.xlsx')\n", "df5.index = pd.to_datetime(df5['date'])\n", "df5.drop('date', axis=1)\n", "df5" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "**resample()** hoitaa myös päällekkäiset aikaleimat. Seuraavassa korvaan päällekkäiset havainnot keskiarvollaan ja puuttuvat havainnot interpoloin." ] }, { "cell_type": "code", "execution_count": 9, "metadata": {}, "outputs": [ { "data": { "text/html": [ "
\n", "\n", "\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
datevalue
date
2019-01-012019-01-01 00:00:002.0
2019-02-012019-02-01 00:00:004.0
2019-03-012019-03-01 00:00:004.0
2019-04-012019-03-31 16:00:005.0
2019-05-012019-05-01 08:00:006.0
2019-06-012019-06-01 00:00:007.0
2019-07-012019-07-01 00:00:008.0
2019-08-012019-08-01 00:00:009.0
\n", "
" ], "text/plain": [ " date value\n", "date \n", "2019-01-01 2019-01-01 00:00:00 2.0\n", "2019-02-01 2019-02-01 00:00:00 4.0\n", "2019-03-01 2019-03-01 00:00:00 4.0\n", "2019-04-01 2019-03-31 16:00:00 5.0\n", "2019-05-01 2019-05-01 08:00:00 6.0\n", "2019-06-01 2019-06-01 00:00:00 7.0\n", "2019-07-01 2019-07-01 00:00:00 8.0\n", "2019-08-01 2019-08-01 00:00:00 9.0" ] }, "execution_count": 9, "metadata": {}, "output_type": "execute_result" } ], "source": [ "df6 = df5.resample('MS').mean().interpolate()\n", "df6" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## Lisätietoa\n", "\n", "**resample()** on monipuolinen toiminto. Lue lisää https://pandas.pydata.org/pandas-docs/dev/reference/api/pandas.DataFrame.resample.html" ] } ], "metadata": { "kernelspec": { "display_name": "Python 3 (ipykernel)", "language": "python", "name": "python3" }, "language_info": { "codemirror_mode": { "name": "ipython", "version": 3 }, "file_extension": ".py", "mimetype": "text/x-python", "name": "python", "nbconvert_exporter": "python", "pygments_lexer": "ipython3", "version": "3.12.2" } }, "nbformat": 4, "nbformat_minor": 4 }