Amaphrojekthi womthombo ovulekile ayi-14 wokuthuthukisa amakhono eSayensi Yedatha (alula, ajwayelekile, aqinile)

Isayensi Yedatha Yabaqalayo

1. Ukuhlaziya Imizwa (Ukuhlaziya isimo ngombhalo)

Buka ukuqaliswa okuphelele kwephrojekthi Yesayensi Yedatha usebenzisa ikhodi yomthombo − I-Sentiment Analysis Project ku-R.

I-Sentiment Analysis iwukuhlaziya amagama ukuze kutholakale imizwa nemibono, okungaba kuhle noma kube kubi. Lolu uhlobo lokuhlukanisa lapho izigaba zingaba kanambambili (ezinhle nezingezinhle) noma zibe ubuningi (ukuthokoza, ukucasuka, ukudabukisa, okubi...). Sizosebenzisa le phrojekthi Yesayensi Yedatha ngo-R futhi sizosebenzisa idathasethi kuphakheji ye-"janeaustenR". Sizosebenzisa izichazamazwi ezijwayelekile ezifana ne-AFINN, i-bing ne-loughran, senze ukujoyina kwangaphakathi futhi ekugcineni sizodala ifu lamagama ukuze sibonise umphumela.

I-albhamu kuphela: R
Isethi yedatha/Iphakheji: janeoustenR

I-athikili ihunyushwe ngosekelo lwe-EDISON Software, okuyiyona yenza amagumbi okulingana okubonakalayo ezitolo zemikhiqizo eminingiFuthi isofthiwe yokuhlola.

2. Ukutholwa Kwezindaba Okungelona iqiniso

Thatha amakhono akho uwayise kwelinye izinga ngokusebenza ku-Data Science Project for Beginners − ukutholwa kwezindaba ezingamanga ngePython.

Izindaba ezingamanga ziwulwazi olungamanga olusakazwa ezinkundleni zokuxhumana nakweminye imidiya ye-inthanethi ukuze kuzuzwe izinhloso zezepolitiki. Kulo mbono wephrojekthi ye-Data Science, sizosebenzisa i-Python ukwakha imodeli engakwazi ukucacisa ngokunembile ukuthi izindaba ziyiqiniso noma zingamanga. Sizodala i-TfidfVectorizer futhi sisebenzise i-PassiveAggressiveClassifier ukuze sihlukanise izindaba zibe "zangempela" kanye "nenkohliso". Sizosebenzisa isethi yedatha yomumo ongu-7796×4 futhi senze yonke into ku-Jupyter Lab.

I-albhamu kuphela: Python

Isethi yedatha/Iphakheji: izindaba.csv

3. Ukuthola Isifo sikaParkinson

Qhubekela phambili ngokusebenza ku-Data Science Project Idea − ukutholwa kwesifo sikaParkinson nge-XGBoost.

Sesiqalile ukusebenzisa i-Data Science ukuze sithuthukise ukunakekelwa kwezempilo nezinsizakalo - uma singabikezela isifo kusenesikhathi, sizoba nezinzuzo eziningi. Ngakho-ke, kulo mbono wephrojekthi Yesayensi Yedatha, sizofunda ukuthi singasithola kanjani isifo sikaParkinson sisebenzisa iPython. Kuyisifo se-neurodegenerative, esiqhubekayo sohlelo lwezinzwa oluphakathi esithinta ukunyakaza futhi sibangele ukuthuthumela nokuqina. Kuthinta ama-neurons akhiqiza i-dopamine ebuchosheni, futhi minyaka yonke, kuthinta abantu abangaphezu kwesigidi esingu-1 eNdiya.

I-albhamu kuphela: Python

Isethi yedatha/Iphakheji: Idatha yedatha ye-UCI ML Parkinsons

Iphrojekthi Yesayensi Yedatha eyinkimbinkimbi ephakathi

4. Ukuqashelwa Kwemizwa Yenkulumo

Bheka ukuqaliswa okugcwele kwesampula yephrojekthi yeSayensi Yedatha − ukuqashelwa kwenkulumo nge-Librosa.

Manje ake sifunde ukusebenzisa imitapo yolwazi ehlukene. Le phrojekthi Yesayensi Yedatha isebenzisa i-librosa ukuze ibone inkulumo. I-SER inqubo yokuhlonza imizwa yomuntu kanye nezimo ezithintekayo enkulumeni. Ngoba sisebenzisa ithoni nephimbo ukuveza imizwelo ngezwi lethu, i-SER ibalulekile. Kodwa njengoba imizwa izimele, isichasiselo somsindo siwumsebenzi onzima. Sizosebenzisa imisebenzi ye-mfcc, chroma ne-mel futhi sisebenzise idathasethi ye-RAVDESS ukuze siqaphele imizwa. Sizodala isihlukanisi se-MLPC sale modeli.

I-albhamu kuphela: Python

Isethi yedatha/Iphakheji: Isethi yedatha ye-RAVDESS

5. Ukutholwa kobulili nobudala

Hlabisa abaqashi umxhwele ngephrojekthi yakamuva ye-Data Science - ukutholwa kobulili nobudala nge-OpenCV.

Lena iSayensi Yedatha ethokozisayo enePython. Usebenzisa isithombe esisodwa kuphela, uzofunda ukubikezela ubulili bomuntu neminyaka yobudala. Kulokhu, sizokwethula ku-Computer Vision nezimiso zayo. Sizokwakha inethiwekhi ye-convolutional neural futhi izosebenzisa amamodeli aqeqeshwe u-Tal Hassner no-Gil Levy kudathasethi ye-Adience. Sizosebenzisa amanye amafayela e-.pb, .pbtxt, .prototxt kanye ne-.caffemodel endleleni.

I-albhamu kuphela: Python

Isethi yedatha/Iphakheji: Ukulalela

6. Ukuhlaziywa Kwedatha ye-Uber

Buka ukuqaliswa okuphelele kwephrojekthi Yesayensi Yedatha ngekhodi yomthombo − Iphrojekthi yokuhlaziya idatha ye-Uber ku-R.

Lena iphrojekthi yokubona idatha ene-ggplot2 lapho sizosebenzisa khona i-R nemitapo yolwazi yayo futhi sihlaziye amapharamitha ahlukahlukene. Sizosebenzisa isethi yedatha ye-Uber Pickups New York futhi sidale ukubonwa kwezikhathi ezihlukene zonyaka. Lokhu kusitshela ukuthi isikhathi siluthinta kanjani uhambo lwamakhasimende.

I-albhamu kuphela: R

Isethi yedatha/Iphakheji: I-Uber Pickups e-New York City dataset

7. Ukutholwa ukozela komshayeli

Thuthukisa amakhono akho ngokusebenza Kuphrojekthi Yesayensi Yedatha Ephezulu - uhlelo lokutholwa kokulala olune-OpenCV & Keras.

Ukushayela ubuthongo kuyingozi kakhulu, cishe kunezingozi eziyinkulungwane minyaka yonke ngenxa yokuzumeka kwabashayeli beshayela. Kule phrojekthi yePython, sizokwakha uhlelo olukwazi ukubona abashayeli abalalayo futhi lubaxwayise ngebhiphu.

Le phrojekthi isetshenziswa kusetshenziswa i-Keras ne-OpenCV. Sizosebenzisa i-OpenCV ukuthola ubuso namehlo futhi ngosizo lwe-Keras sizohlukanisa isimo seso (Ivuliwe noma Ivaliwe) sisebenzisa izindlela ezijulile zenethiwekhi ye-neural.

8. Ingxoxo

Yakha i-chatbot nePython futhi uthathe igxathu eliya phambili emsebenzini wakho - I-Chatbot ne-NLTK & Keras.

Ama-Chatbots ayingxenye ebalulekile yebhizinisi. Amabhizinisi amaningi kufanele anikeze izinsizakalo kumakhasimende awo futhi kuthatha amandla amaningi, isikhathi nomzamo ukuwasebenzela. Ama-Chatbot angakwazi ukwenza okuningi kokusebenzisana kwamakhasimende ngokuzenzakalelayo ngokuphendula eminye yemibuzo evamile ebuzwa ngamakhasimende. Kunezinhlobo ezimbili ze-chatbot: Isizinda esicacisiwe kanye nesizinda esivulekile. I-chatbot eqondene nesizinda ivamise ukusetshenziselwa ukuxazulula inkinga ethile. Ngakho-ke, udinga ukuyenza ngokwezifiso ukuze isebenze ngempumelelo ensimini yakho. Ama-chatbot esizinda esivulekile angabuzwa noma yimiphi imibuzo, ngakho ukuwaqeqesha kudinga inani elikhulu ledatha.

Idatha isethi: Inhloso yefayela le-json

I-albhamu kuphela: Python

Amaphrojekthi we-Advanced Data Science

9. Ijeneretha yamagama-ncazo wesithombe

Bheka ukuqaliswa okuphelele kwephrojekthi ngekhodi yomthombo − Ijeneretha yamagama-ncazo wesithombe ene-CNN ne-LSTM.

Ukuchaza okusesithombeni kuwumsebenzi olula kubantu, kodwa kumakhompyutha, isithombe siyiqoqo lezinombolo ezimele inani lombala wephikseli ngayinye. Lona umsebenzi onzima wamakhompyutha. Ukuqonda ukuthi yini esesithombeni bese udala incazelo yolimi lwemvelo (isib. IsiNgisi) omunye umsebenzi onzima. Le phrojekthi isebenzisa amasu okufunda ajulile lapho sisebenzisa khona i-Convolutional Neural Network (CNN) ene-Recurrent Neural Network (LSTM) ukuze sakhe ijeneretha yencazelo yesithombe.

Idatha isethi: I-Flickr 8K

I-albhamu kuphela: Python

Uhlaka: UKeras

10. Ukutholwa Kokukhwabanisa Kwekhadi Lesikweletu

Yenza okusemandleni akho ngokusebenza ngombono wephrojekthi ye-Data Science − ukutholwa kokukhwabanisa kwekhadi lesikweletu ngokufunda komshini.

Manje usuqalile ukuqonda izindlela nemibono. Asiqhubekele kumaphrojekthi wesayensi yedatha ethuthukisiwe. Kule phrojekthi, sizosebenzisa ulimi lwe-R olunama-algorithms afana nala izihlahla zesinqumo, ukuhlehla kwezinto, amanethiwekhi okwenziwa kwemizwa kanye nesihlukanisi esithuthukisa ukuthambeka. Sizosebenzisa isethi yedatha yemisebenzi yekhadi ukuze sihlukanise okwenziwa ngekhadi lesikweletu njengokuwumgunyathi nokwangempela. Sizozikhethela amamodeli ahlukene futhi sakhe amajika okusebenza.

I-albhamu kuphela: R

Isethi yedatha/Iphakheji: Isethi yedatha Yokwenziwe Ngekhadi

11. Uhlelo lokuncoma i-Movie

Hlola ukuqaliswa kwephrojekthi ye-Data Science engcono kakhulu nge-Source Code - I-Movie Recommendation System ku-R

Kule phrojekthi Yesayensi Yedatha, sizosebenzisa u-R ukuze senze izincomo ze-movie ngokufunda ngomshini. Isistimu yokuncoma ithumela iziphakamiso kubasebenzisi ngenqubo yokuhlunga ngokusekelwe kulokho okuthandwa ngabanye abasebenzisi kanye nomlando wokuphequlula. Uma u-A no-B bethanda i-Home Alone, futhi u-B ethanda i-Mean Girls, ungaphakamisa u-A - bangase bayithande futhi. Lokhu kuvumela amaklayenti ukuthi ahlanganyele nenkundla.

I-albhamu kuphela: R

Isethi yedatha/Iphakheji: Isethi yedatha ye-MovieLens

12. Ukuhlukaniswa Kwekhasimende

Gcoba abaqashi umxhwele ngephrojekthi ye-Data Science (kuhlanganise nekhodi yomthombo) - Ukuhlukaniswa kwekhasimende ngokufunda komshini.

Ukuhlukaniswa komthengi kuwuhlelo lokusebenza oludumile ukufunda okungajwayelekile. Ngokusebenzisa ukuhlanganisa, izinkampani zichaza izingxenye zamakhasimende ukuze zisebenze nesizinda somsebenzisi esingaba khona. Bahlukanisa amakhasimende ngamaqembu ngokuvumelana nezici ezivamile ezinjengobulili, ubudala, izinto abazithandayo, nemikhuba yokuchitha imali, ukuze bakwazi ukumaketha ngempumelelo imikhiqizo yabo eqenjini ngalinye. Sizosebenzisa I-K-isho ukuhlangana, kanye nokubona ngeso lengqondo ukusatshalaliswa ngokobulili nobudala. Sibe sesihlaziya amazinga abo onyaka wemali engenayo kanye nezindleko.

I-albhamu kuphela: R

Isethi yedatha/Iphakheji: Isethi yedatha ye-Mall_Customers

13. Ukuhlukaniswa Komdlavuza Webele

Bona ukuqaliswa okuphelele kwephrojekthi ye-Data Science ku-Python − Ukwahlukaniswa Komdlavuza Webele Ngokusebenzisa Ukufunda Okujulile.

Ukubuyela emnikelweni wezokwelapha wesayensi yedatha, ake sifunde ukuthi ungawuthola kanjani umdlavuza webele ngePython. Sizosebenzisa i-IDC_regular dataset ukuze sithole invasive ductal carcinoma, uhlobo oluvame kakhulu lomdlavuza webele. Ikhula emiseleni yobisi, ingene kuzicubu ze-fibrous noma amafutha e-mammary gland ngaphandle komgudu. Kulo mbono wephrojekthi yesayensi yokuqoqwa kwedatha, sizowusebenzisa Ukufunda Okujulile kanye nomtapo wezincwadi wakwaKeras ukuze uhlukaniswe.

I-albhamu kuphela: Python

Isethi yedatha/Iphakheji: I-IDC_evamile

14. Ukuqashelwa Kwezimpawu Zethrafikhi

Ukuzuza ukunemba kubuchwepheshe bezimoto ezizishayelayo nge-Data Science project on ukuqashelwa kwezimpawu zethrafikhi kusetshenziswa i-CNN umthombo ovulekile.

Izimpawu zomgwaqo kanye nemithetho yomgwaqo ibaluleke kakhulu kuwo wonke umshayeli ukugwema izingozi. Ukuze ulandele umthetho, okokuqala udinga ukuqonda ukuthi uphawu lomgwaqo lubukeka kanjani. Umuntu kufanele afunde zonke izimpawu zomgwaqo ngaphambi kokuba anikezwe ilungelo lokushayela noma iyiphi imoto. Kodwa manje inani lezimoto ezizimele liyakhula, futhi esikhathini esizayo esiseduze, umuntu ngeke esakwazi ukushayela imoto yedwa. Kuphrojekthi Yokuqaphela Uphawu Lomgwaqo, uzofunda ukuthi uhlelo lungabona kanjani uhlobo lophawu lomgwaqo ngokuthatha isithombe njengokufaka. I-German Road Sign Recognition Reference Dataset (GTSRB) isetshenziselwa ukwakha inethiwekhi ye-neural ejulile ukuze ibone isigaba okuyingxenye yaso uphawu. Futhi sakha i-GUI elula yokuxhumana nohlelo lokusebenza.

I-albhamu kuphela: Python

Idatha isethi: I-GTRB (Ibhentshimakhi yokuqaphela uphawu lwethrafikhi yaseJalimane)

Funda kabanzi

• 52 amasethi edatha amaphrojekthi okuqeqesha
• I-Dojo yangaphambili: amaphrojekthi okuqeqesha amakhono onjiniyela (ama-5 amasha + ama-43 amadala)
• I-12 ephezulu ethakazelisa kakhulu ye-IT Dynamic Infographics

Source: www.habr.com