Kuongororwa kweMusiyano uye Kutsauka Kwakajairwa muKugoverwa kweData
Muhuwandu hwedata, kunzwisisa kugoverwa kwedata kwakakosha sekunzwisisa kukosha kwepakati senge mean kana median. Maseti maviri edata anogona kuva neavhareji yakafanana, asi kugoverwa kwawo kwakasiyana zvikuru: rimwe rinogona kunge rakaungana zvakabatana neavhareji, nepo rimwe richipararira zvakanyanya. Apa ndipo panopinda musiyano uye kutsauka kwakajairwa—ndizvo zvinonyanya kukosha pakuyera kuti data rakasiyana sei nepakati paro. Chinyorwa chino chinokurukura pfungwa dzavo, mafomula, dudziro, uye mienzaniso yekushandiswa kwavo mukuongorora data.
1. Sei Kuparadzira Data Kwakakosha?
Kupararira kwedata kunopa ruzivo nezvekuenderana uye njodzi. Semuenzaniso, kana tichitarisa mamakisi ebvunzo, avhareji yemakirasi A neB inogona kuva 80. Zvisinei, kana musiyano wemakisi ekirasi A uri mudiki, ruzhinji rwevadzidzi runoita zvakafanana. Kusiyana neizvi, kana musiyano wemakisi ekirasi B wakakura, zvingangoita kuti vamwe vadzidzi vane mamakisi akakwira uye vamwe vane mamakisi akaderera. Mubhizinesi, kupararira kwedata rekutengesa kunoratidza kugadzikana kwemari; mune zvemari, kupararira kwemari inobuda kunoratidza mwero wenjodzi.
Nekunzwisisa kusiyana uye kutsauka kwakajairwa, vanoita sarudzo vanogona:
- Ongorora kana maitiro acho akagadzikana kana kuti kwete (semuenzaniso kugadzirwa kwefekitori).
- Kuenzanisa kuenderana pakati pemapoka (semuenzaniso nzira mbiri dzekudzidza).
- Kuziva ruzivo rusina kukosha rwakakodzera kuongororwa.
- Kufungidzira kusava nechokwadi mune fungidziro uye mamodheru.
2. Pfungwa Yekutanga Yekusiyana Kwemashoko
Kusiyana kunoyera avhareji yekutsauka kwese kweseti yedata kubva paavhareji. Kutsauka ndiko musiyano uripo pakati pehuwandu hwedata neavhareji. Kana huwandu hwakawanda huri kure neavhareji, musiyano uchava mukuru. Kana huwandu huri pedyo neavhareji, musiyano uchava mudiki.
Ngatitii pane data: \(x_1, x_2, …, x_n\) ine avhareji ye \(\bar{x}\). Kutsauka kwedata rega rega ndiko \(x_i – \bar{x}\). Zvisinei, kana kutsauka kwakawedzerwa zvakananga, mhedzisiro inogara iri zero nekuti kune kutsauka kwakanaka uye kusina kunaka kunodzimana. Kuti izvi zvikunde, kutsauka kwacho kunoiswa sikweya kuitira kuti zvese zvive zvakanaka. Apa ndipo panozvarwa musiyano.
a) Kusiyana kweVanhu
Kana data racho richionekwa serinomiririra huwandu hwevanhu vose, musiyano wehuwandu hwevanhu unonyorwa seizvi:
\[
\sigma^2 = \frac{\sum_{i=1}^{N}(x_i – \mu)^2}{N}
\]
di mana:
– \(N\) ndiyo nhamba yedata revanhu,
– \(\mu\) iavhareji yevanhu,
– \(\sigma^2\) ndiko kusiyana kwehuwandu hwevanhu.
b) Musiyano wemuenzaniso
Kana data iri sampuli kubva kuvanhu vakawanda, musiyano wemuenzaniso unoshandiswa:
\[
s^2 = \frac{\sum_{i=1}^{n}(x_i – \bar{x})^2}{n-1}
\]
Chikamu chinopatsanura \(n-1\) chinonzi Bessel correction , uye chinoshandiswa kuona kuti kufungidzira kwevariance yehuwandu hwevanhu hakuna rusaruro. Chaizvoizvo, nekuti sampuli mean inoverengwa kubva padata pachayo, pane "kurasikirwa kwemadhigirii erusununguko," saka chikamu chinopatsanura chinogadziriswa zvinoenderana.
3. Kutsauka Kwakajairwa: Mudzi weKusiyana
Kusiyana-siyana kune dambudziko rimwe chete rinoshanda: mayuniti ayo ndiwo sikweya yemayuniti edata. Kana data riri mu "rupiah", kusiyana kuri mu "rupiah²", izvo zvakaoma kududzira zvakananga. Saka, tinoshandisa standard deviation, inova square root ye kusiyana.
a) Kutsauka Kwehuwandu Hwevanhu
\[
\sigma = \sqrt{\sigma^2}
\]
b) Muenzaniso weKutsauka Kwemuyero
\[
s = \sqrt{s^2}
\]
Kutsauka kwakajairika kune mayuniti akafanana nedata rekutanga, zvichiita kuti zvive nyore kunzwisisa. Kutsauka kwakajairika kunoratidza data rakapararira zvakanyanya; kutsauka kwakaderera kunoratidza seti yedata yakamanikana.
4. Muenzaniso weKuverenga Uri Nyore
Semuenzaniso, data rezvibodzwa zvebvunzo: 70, 75, 80, 85, 90.
1) Verenga avhareji:
\[
\bar{x} = \frac{70+75+80+85+90}{5} = 80
\]
2) Verenga kutsauka kwemutengo wega wega kubva paavhareji:
– 70: \(70-80=-10\)
– 75: \(75-80=-5\)
– 80: \(80-80=0\)
– 85: \(85-80=5\)
– 90: \(90-80=10\)
3) Kusiyanisa pakati pemativi maviri:
– 100, 25, 0, 25, 100
4) Wedzera:
\[
\sum (x_i-\bar{x})^2 = 250
\]
5) Musiyano wemuenzaniso:
\[
s^2 = \frac{250}{5-1} = 62.5
\]
6) Muenzaniso wekutsauka kwakajairika:
\[
s = \sqrt{62.5} \inenge 7.91
\]
Dudziro: avhareji yemapoinzi i80, uye "kazhinji" mamakisi anosiyana nemapoinzi angangoita 7-8 kubva paavhareji.
5. Kududzirwa kweMusiyano uye Kutsauka Kwakajairwa
Kusiyana-siyana uye kutsauka kwakajairwa hazvisi nhamba chete; zvinofanira kutsanangurwa zvichienderana nemamiriro ezvinhu.
– Kutsauka kudiki: kuenderana kwakanyanya. Semuenzaniso, maitiro ekugadzira ane kutsauka kudiki kwazvo muhukuru hwechigadzirwa anoratidza hunhu hwakagadzikana.
– Kuchinja-chinja kukuru: kuchinja-chinja kukuru. Mukudyara mari, kuchinja-chinja kukuru kwemubereko zvinoreva kusagadzikana kukuru (njodzi huru).
- Kuenzanisa mapoka: kana mapoka maviri aine avhareji imwe chete asi akasiyana, boka rine kutsauka kudiki rinofanana zvakanyanya.
Zvisinei, zvakakosha kuyeuka kuti standard deviation inonyatsoenderana nezvinhu zvisingawanikwe. Kukosha kumwe chete kwakanyanya kunogona kuwedzera zvakanyanya variance uye standard deviation. Nokudaro, ongororo yekugoverwa inowanzo wedzerwa ne visualizations (histograms, boxplots) kana zviyero zvakasimba zvakaita seIQR (interquartile range).
6. Hukama neKugoverwa Kwakajairika uye Mitemo yeEmpirical
Muchikamu che "normal distribution" (bell curve), "standard deviation" ine chirevo chakasimba. Pane mutemo we "empirical" unowanzoshandiswa:
– Inenge 68% yedata iri muhuwandu \(\bar{x} \pm 1s\)
– Inenge 95% yedata iri muhuwandu \(\bar{x} \pm 2s\)
– Inenge 99,7% yedata iri muhuwandu \(\bar{x} \pm 3s\)
Mutemo uyu unobatsira kuita tsananguro nekukurumidza, semuenzaniso kuongorora kana kukosha "kusiri kwechisikigo" kana kuti kuchiri mukati mehuwandu hwakajairika.
7. Mashandisirwo muMinda Yakasiyana-siyana
1) Dzidzo: Kutarisa kugoverwa kwemagiredhi evadzidzi. Kusawirirana kudiki kunoratidza mhedzisiro yekudzidza yakaenzana, nepo kusawirirana kukuru kuchiratidza kusanzwisisana.
2) Indasitiri: kudzora mhando. Kusiyana kunoshandiswa kuongorora kuenderana kwekugadzirwa.
3) Zvemari: zvinoyera kusagadzikana kwemitengo yemasheya, kudzoka kwemari inowanikwa muzvitoro, uye njodzi yekudyara mari.
4) Hutano: kuona kusiyana kweBP, huwandu hweshuga, kana zvimwe zviratidzo zvekiriniki muvarwere.
5) Tsvagiridzo yemagariro evanhu: kuongorora kusiyana kwemhinduro dzeongororo uye kusiyana kwehunhu hwevapinduri.
8. Zvikanganiso Zvakajairika uye Mazano Anoshanda
Zvimwe zvikanganiso zvakajairika:
– Kushandisa musiyano wemuenzaniso (divisor \(n-1\)) kunyangwe data riri iro huwandu hwakazara, kana kuti zvinopesana.
- Dudzira musiyano usingatarise mayuniti ayo emativi mana; zvakachengeteka kushandisa kutsauka kwakajairwa pakududzira.
- Usafuratira zvinhu zvisingawanzoitiki; zvakanaka kutanga watarisa data racho.
– Enzanisa kutsauka kwakajairwa pakati pedata nezviyero zvakasiyana pasina kurongeka; mune dzimwe nguva, shandisa coefficient of variation (CV) kureva \(CV = \frac{s}{\bar{x}}\times 100\%\) kuti uwane kuenzanisa kwakaringana.
Penutup
Kusiyana-siyana uye kutsauka kwakajairwa zvishandiso zvakakosha pakunzwisisa kugoverwa kwedata. Kusiyana-siyana kunopa hwaro hwakasimba hwemasvomhu, nepo kutsauka kwakajairwa kuchipa chiyero chiri nyore kududzira nekuti chakafanana nedata rekutanga. Nekushandisa zviyero izvi zviviri, tinogona kunyatsoongorora kuenderana, njodzi, uye mutsauko muhunhu hwekugoverwa pakati pemaseti edata. Mukudzidzira kwekuongorora data, kusiyana-siyana uye kutsauka kwakajairwa zvinoshandiswa zvakanyanya pamwe chete nezviyero zvepakati uye kuona kuti data racho rionekwe sei uye kuita sarudzo dzine ruzivo rwakawanda.
Kana muchida, ndinogona kuwedzera mienzaniso yakaoma yekuverenga (semuenzaniso data rakarongwa), kana kutsanangura hukama hwe standard deviation ne z-score uye outlier detection.
