๐Ÿ“‡ Machine Learning

๐Ÿ“‡ Machine Learning

[ML] Hyperparameter Tuning (ํ•˜์ดํผ ํŒŒ๋ผ๋ฏธํ„ฐ ํŠœ๋‹)

๋จธ์‹ ๋Ÿฌ๋‹ ๋ชจ๋ธ์„ ํ•™์Šตํ• ๋•Œ ์ค‘์š”ํ•œ ์š”์†Œ์ค‘ ํ•˜๋‚˜์ธ Hyperparamter(ํ•˜์ดํผ ํŒŒ๋ผ๋ฏธํ„ฐ)์— ๋ฐํ•˜์—ฌ ์•Œ์•„๋ณด๊ฒ ์Šต๋‹ˆ๋‹ค.Hyperparameter? (ํ•˜์ดํผ ํŒŒ๋ผ๋ฏธํ„ฐ๋ž€?)ํ•˜์ดํผ ํŒŒ๋ผ๋ฏธํ„ฐ๋Š” ๋จธ์‹ ๋Ÿฌ๋‹ ๋ชจ๋ธ์„ ํ•™์Šตํ•˜๊ธฐ ์ „์— ์„ค์ •ํ•ด์•ผ ํ•˜๋Š” ๊ฐ’์œผ๋กœ, ํ•™์Šต ๊ณผ์ • ์ค‘์—๋Š” ๋ณ€๊ฒฝ๋˜์ง€ ์•Š์Šต๋‹ˆ๋‹ค.์ด๋Š” ๋ชจ๋ธ์˜ ์„ฑ๋Šฅ๊ณผ ํ•™์Šต ์†๋„์— ํฐ ์˜ํ–ฅ์„ ๋ฏธ์น˜๋Š” ์ค‘์š”ํ•œ ์š”์†Œ์ž…๋‹ˆ๋‹ค. Hyperparameter ์˜ˆ์‹œ๊ฒฐ์ • ํŠธ๋ฆฌ์˜ ์ตœ๋Œ€ ๊นŠ์ด: ํŠธ๋ฆฌ๊ฐ€ ์–ผ๋งˆ๋‚˜ ๊นŠ๊ฒŒ ์„ฑ์žฅํ•  ์ˆ˜ ์žˆ๋Š”์ง€๋ฅผ ๊ฒฐ์ •ํ•˜๋ฉฐ, ๋ชจ๋ธ์˜ ๋ณต์žก์„ฑ์„ ์กฐ์ ˆํ•ฉ๋‹ˆ๋‹ค.SVM์˜ ์ปค๋„ ์ข…๋ฅ˜: Support Vector Machine์—์„œ ์‚ฌ์šฉํ•˜๋Š” ์ปค๋„์˜ ์ข…๋ฅ˜๋ฅผ ์„ค์ •ํ•˜์—ฌ, ๋ฐ์ดํ„ฐ๋ฅผ ๋ณ€ํ™˜ํ•˜๋Š” ๋ฐฉ๋ฒ•์„ ์ •์˜ํ•ฉ๋‹ˆ๋‹ค.์‹ ๊ฒฝ๋ง์˜ ํ•™์Šต๋ฅ : ์‹ ๊ฒฝ๋ง์—์„œ ๊ฐ€์ค‘์น˜๋ฅผ ์—…๋ฐ์ดํŠธํ•  ๋•Œ ์‚ฌ์šฉํ•˜๋Š” ํ•™์Šต๋ฅ ์€ ๋ชจ๋ธ์˜ ์ˆ˜๋ ด ์†๋„์™€ ํ•™์Šต ํ’ˆ์งˆ์— ์˜..

๐Ÿ“‡ Machine Learning

[ML] Reinforcement Learning (๊ฐ•ํ™” ํ•™์Šต) - Q-Learning

๊ฐ•ํ™” ํ•™์Šต (Reinforcement Learning) ์ด๋ž€?๊ฐ•ํ™” ํ•™์Šต์€ ์—์ด์ „ํŠธ๊ฐ€ ํ™˜๊ฒฝ๊ณผ ์ƒํ˜ธ์ž‘์šฉํ•˜๋ฉด์„œ ๋ณด์ƒ์„ ์ตœ๋Œ€ํ™”ํ•˜๋Š” ํ–‰๋™ ์ •์ฑ…์„ ํ•™์Šตํ•˜๋Š” ๋ฐฉ๋ฒ•์ž…๋‹ˆ๋‹ค.์—์ด์ „ํŠธ๋Š” ์ฃผ์–ด์ง„ ํ™˜๊ฒฝ์—์„œ ์ตœ์ ์˜ ํ–‰๋™์„ ํ•™์Šตํ•˜์—ฌ ์žฅ๊ธฐ์ ์œผ๋กœ ๋ˆ„์  ๋ณด์ƒ์„ ์ตœ๋Œ€ํ™”ํ•˜๋Š” ๊ฒƒ์„ ๋ชฉํ‘œ๋กœ ํ•ฉ๋‹ˆ๋‹ค.๊ฐ•ํ™” ํ•™์Šต์˜ ๋ชฉ์ ์ตœ์ ์˜ ํ–‰๋™ ์ •์ฑ… ํ•™์Šต: ์—์ด์ „ํŠธ๊ฐ€ ์ฃผ์–ด์ง„ ํ™˜๊ฒฝ์—์„œ ์ตœ์ ์˜ ํ–‰๋™์„ ์„ ํƒํ•˜์—ฌ ๋ˆ„์  ๋ณด์ƒ์„ ์ตœ๋Œ€ํ™”ํ•˜๋Š” ์ •์ฑ…์„ ํ•™์Šตํ•˜๋Š” ๊ฒƒ์ด ๋ชฉ์ ์ž…๋‹ˆ๋‹ค.Q-learning๊ฐ•ํ™”ํ•™์Šต์—์„œ, Q-learning์ด๋ผ๋Š” ๋ฐฉ๋ฒ•์ด ์žˆ์Šต๋‹ˆ๋‹ค. ํ•œ๋ฒˆ ์ž์„ธํžˆ ์•Œ์•„๋ณด๊ฒ ์Šต๋‹ˆ๋‹ค. Q-learning์€ ์ƒํƒœ-ํ–‰๋™ ๊ฐ€์น˜ ํ•จ์ˆ˜(Q-ํ•จ์ˆ˜)๋ฅผ ํ•™์Šตํ•˜์—ฌ ์ตœ์ ์˜ ์ •์ฑ…์„ ์ฐพ๋Š” ๊ฐ•ํ™” ํ•™์Šต ๋ฐฉ๋ฒ• ์ค‘ ํ•˜๋‚˜์ž…๋‹ˆ๋‹ค.์ด ๋ฐฉ๋ฒ•์€ ์ฃผ์–ด์ง„ ์ƒํƒœ์—์„œ ์–ด๋–ค ํ–‰๋™์„ ์ทจํ•ด์•ผ ํ•˜๋Š”์ง€๋ฅผ ๊ฒฐ์ •ํ•˜๋Š” ๋ฐ ์‚ฌ์šฉ๋ฉ๋‹ˆ๋‹ค. Q-le..

๐Ÿ“‡ Machine Learning

[ML] Recommender System (์ถ”์ฒœ์‹œ์Šคํ…œ)

์ด๋ฒˆ์—๋Š” Recommend System (์ถ”์ฒœ์‹œ์Šคํ…œ)์— ๋ฐํ•˜์—ฌ ์•Œ์•„๋ณด๊ฒ ์Šต๋‹ˆ๋‹ค. ์ถ”์ฒœ ์‹œ์Šคํ…œ์€ ์‚ฌ์šฉ์ž์™€ ์•„์ดํ…œ ๊ฐ„์˜ ๊ด€๊ณ„๋ฅผ ๋ถ„์„ํ•˜์—ฌ ์‚ฌ์šฉ์ž์—๊ฒŒ ์ ํ•ฉํ•œ ์•„์ดํ…œ์„ ์ถ”์ฒœํ•˜๋Š” ์‹œ์Šคํ…œ์ž…๋‹ˆ๋‹ค.์ด๋Ÿฌํ•œ ์‹œ์Šคํ…œ์€ ๋‹ค์–‘ํ•œ ๋ฐ์ดํ„ฐ๋ฅผ ํ™œ์šฉํ•ด ์‚ฌ์šฉ์ž์—๊ฒŒ ๋งž์ถคํ˜• ์ถ”์ฒœ์„ ์ œ๊ณตํ•˜๋ฉฐ, ๋‹ค์–‘ํ•œ ๋ถ„์•ผ์—์„œ ๋„๋ฆฌ ์‚ฌ์šฉ๋ฉ๋‹ˆ๋‹ค.์ถ”์ฒœ ์‹œ์Šคํ…œ์˜ ๋ชฉ์  ์‚ฌ์šฉ์ž ๋งŒ์กฑ๋„ ํ–ฅ์ƒ: ์‚ฌ์šฉ์ž๊ฐ€ ์„ ํ˜ธํ•  ๋งŒํ•œ ์•„์ดํ…œ์„ ์ถ”์ฒœํ•˜์—ฌ ๋งŒ์กฑ๋„๋ฅผ ๋†’์ด๊ณ  ์‚ฌ์šฉ์ž ๊ฒฝํ—˜์„ ๊ฐœ์„ ํ•ฉ๋‹ˆ๋‹ค.ํŒ๋งค ์ฆ๋Œ€: ์ ์ ˆํ•œ ์ œํ’ˆ ์ถ”์ฒœ์„ ํ†ตํ•ด ๊ตฌ๋งค๋ฅผ ์ด‰์ง„ํ•˜๊ณ , ๋งค์ถœ์„ ์ฆ๋Œ€์‹œํ‚ต๋‹ˆ๋‹ค.์‚ฌ์šฉ์ž ์ฐธ์—ฌ ์ฆ๋Œ€: ๋งž์ถคํ˜• ์ฝ˜ํ…์ธ ๋ฅผ ์ œ๊ณตํ•˜์—ฌ ์‚ฌ์šฉ์ž๊ฐ€ ๋” ์ž์ฃผ, ๋” ์˜ค๋ž˜ ์„œ๋น„์Šค๋ฅผ ์ด์šฉํ•˜๋„๋ก ์œ ๋„ํ•ฉ๋‹ˆ๋‹ค.์ถ”์ฒœ ์‹œ์Šคํ…œ์˜ ์ข…๋ฅ˜์ถ”์ฒœ ์‹œ์Šคํ…œ์˜ ์ข…๋ฅ˜๋Š” ์–ด๋– ํ•œ ์ข…๋ฅ˜๋“ค์ด ์žˆ์„๊นŒ์š”? ํ•œ๋ฒˆ ์•Œ์•„๋ณด๊ฒ ์Šต๋‹ˆ๋‹ค. 3๊ฐ€์ง€ ์ข…๋ฅ˜๊ฐ€ ์žˆ์Šต๋‹ˆ๋‹ค. ์ž์„ธํžˆ ํ•œ๋ฒˆ ..

๐Ÿ“‡ Machine Learning

[ML] Emsemble Methods (์•™์ƒ๋ธ” ๊ธฐ๋ฒ•)

์ด๋ฒˆ์—” Emsemble Methods (์•™์ƒ๋ธ” ๊ธฐ๋ฒ•)์— ๋ฐํ•˜์—ฌ ํ•œ๋ฒˆ ์•Œ์•„๋ณด๊ฒ ์Šต๋‹ˆ๋‹ค. ์•™์ƒ๋ธ” ๊ธฐ๋ฒ•์€ ์—ฌ๋Ÿฌ ๊ฐœ์˜ ์˜ˆ์ธก ๋ชจ๋ธ์„ ๊ฒฐํ•ฉํ•˜์—ฌ ๋‹จ์ผ ๋ชจ๋ธ๋ณด๋‹ค ๋” ๋‚˜์€ ์„ฑ๋Šฅ์„ ์–ป๋Š” ๋ฐฉ๋ฒ•์ž…๋‹ˆ๋‹ค.์ด๋ฅผ ํ†ตํ•ด ์˜ˆ์ธก์˜ ์ •ํ™•๋„๋ฅผ ๋†’์ด๊ณ , ๋ชจ๋ธ์˜ ์•ˆ์ •์„ฑ์„ ํ–ฅ์ƒ์‹œํ‚ค๋ฉฐ, ๊ณผ์ ํ•ฉ์„ ์ค„์ผ ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.Emsemble ๊ธฐ๋ฒ•์˜ ๋ชฉ์ ์•™์ƒ๋ธ” ๊ธฐ๋ฒ•์„ ์‚ฌ์šฉํ•˜๋Š” ๋ชฉ์ ์€ ๊ณผ์—ฐ ๋ฌด์—‡์ผ๊นŒ์š”?  ์˜ˆ์ธก ์„ฑ๋Šฅ ํ–ฅ์ƒ: ์—ฌ๋Ÿฌ ๋ชจ๋ธ์„ ๊ฒฐํ•ฉํ•˜์—ฌ ๊ฐœ๋ณ„ ๋ชจ๋ธ๋ณด๋‹ค ๋” ๋†’์€ ์˜ˆ์ธก ์ •ํ™•๋„๋ฅผ ๋‹ฌ์„ฑํ•ฉ๋‹ˆ๋‹ค.๊ณผ์ ํ•ฉ ๊ฐ์†Œ: ๋‹ค์–‘ํ•œ ๋ชจ๋ธ์˜ ๊ฒฐ๊ณผ๋ฅผ ๊ฒฐํ•ฉํ•จ์œผ๋กœ์จ ๊ฐœ๋ณ„ ๋ชจ๋ธ์ด ํ•™์Šต ๋ฐ์ดํ„ฐ์— ๊ณผ์ ํ•ฉ๋˜๋Š” ๊ฒƒ์„ ๋ฐฉ์ง€ํ•ฉ๋‹ˆ๋‹ค.์•ˆ์ •์„ฑ ํ–ฅ์ƒ: ๋ชจ๋ธ์˜ ๋ณ€๋™์„ฑ์„ ์ค„์ด๊ณ  ์˜ˆ์ธก์˜ ์ผ๊ด€์„ฑ์„ ๋†’์ด๋Š” ๋ฐ ๋„์›€์„ ์ค๋‹ˆ๋‹ค.Emsemble ๊ธฐ๋ฒ•์˜ ์ข…๋ฅ˜์•™์ƒ๋ธ” ๊ธฐ๋ฒ•์€ 3๊ฐ€์ง€์˜ ์ข…๋ฅ˜๊ฐ€ ์žˆ์Šต๋‹ˆ๋‹ค. ์•„๋ž˜์—์„œ ์ž์„ธํžˆ ํ•œ๋ฒˆ ์„ค๋ช…ํ•ด ๋ณด..

๐Ÿ“‡ Machine Learning

[ML] ์—ฐ๊ด€ ๊ทœ์น™ ํ•™์Šต (Association Rule Learning)

์ด๋ฒˆ์—๋Š” ์—ฐ๊ด€ ๊ทœ์น™ ํ•™์Šต (Association Rule Learning)์— ๋ฐํ•˜์—ฌ ํ•œ๋ฒˆ ์•Œ์•„๋ณด๊ฒ ์Šต๋‹ˆ๋‹ค. ์—ฐ๊ด€ ๊ทœ์น™ ํ•™์Šต์€ ๋ฐ์ดํ„ฐ๋ฒ ์ด์Šค์—์„œ ํ•ญ๋ชฉ ๊ฐ„์˜ ํฅ๋ฏธ๋กœ์šด ๊ด€๊ณ„๋ฅผ ์ฐพ๋Š” ๋น„์ง€๋„ ํ•™์Šต ๋ฐฉ๋ฒ•์ž…๋‹ˆ๋‹ค.์ด ๋ฐฉ๋ฒ•์€ ํŠน์ • ํ•ญ๋ชฉ์ด ๋‚˜ํƒ€๋‚  ๋•Œ ๋‹ค๋ฅธ ํ•ญ๋ชฉ์ด ํ•จ๊ป˜ ๋‚˜ํƒ€๋‚  ํ™•๋ฅ ์„ ๊ณ„์‚ฐํ•˜์—ฌ ์œ ์šฉํ•œ ํŒจํ„ด์„ ๋„์ถœํ•˜๋Š” ๋ฐ ์‚ฌ์šฉ๋ฉ๋‹ˆ๋‹ค.์ฃผ๋กœ Apriori ์•Œ๊ณ ๋ฆฌ์ฆ˜๊ณผ FP-Growth ์•Œ๊ณ ๋ฆฌ์ฆ˜์ด ๋„๋ฆฌ ์‚ฌ์šฉ๋ฉ๋‹ˆ๋‹ค.์—ฐ๊ด€ ๊ทœ์น™ ํ•™์Šต (Association Rule Learning)์˜ ํŠน์ง•์—ฐ๊ด€ ๊ทœ์น™ ํ•™์Šต์— ํŠน์ง•๋“ค์— ๋ฐํ•˜์—ฌ ์•Œ์•„๋ณด๊ฒ ์Šต๋‹ˆ๋‹ค. ๋นˆ๋ฐœ ํ•ญ๋ชฉ ์ง‘ํ•ฉ ๊ธฐ๋ฐ˜: ์—ฐ๊ด€ ๊ทœ์น™ ํ•™์Šต์€ ๋นˆ๋ฐœ ํ•ญ๋ชฉ ์ง‘ํ•ฉ์„ ๊ธฐ๋ฐ˜์œผ๋กœ ์œ ์˜๋ฏธํ•œ ์—ฐ๊ด€ ๊ทœ์น™์„ ๋„์ถœํ•ฉ๋‹ˆ๋‹ค. ๋นˆ๋ฐœ ํ•ญ๋ชฉ ์ง‘ํ•ฉ์€ ์ผ์ • ๋นˆ๋„ ์ด์ƒ ๋‚˜ํƒ€๋‚˜๋Š” ํ•ญ๋ชฉ๋“ค์˜ ์ง‘ํ•ฉ์„ ์˜๋ฏธํ•ฉ๋‹ˆ๋‹ค.๊ณ„์‚ฐ ํšจ์œจ์„ฑ: Apriori์™€ FP-..

๐Ÿ“‡ Machine Learning

[ML] t-SNE (t-Distributed Stochastic Neighbor Embedding)

์ด๋ฒˆ์—” t-SNE (t-Distributed Stochastic Neighbor Embedding)์— ๋ฐํ•˜์—ฌ ํ•œ๋ฒˆ ์•Œ์•„๋ณด๊ฒ ์Šต๋‹ˆ๋‹ค. t-SNE๋Š” ๊ณ ์ฐจ์› ๋ฐ์ดํ„ฐ๋ฅผ ์ €์ฐจ์› ๊ณต๊ฐ„์— ํšจ๊ณผ์ ์œผ๋กœ ์‹œ๊ฐํ™”ํ•˜๊ธฐ ์œ„ํ•ด ๊ฐœ๋ฐœ๋œ ๋น„์„ ํ˜• ์ฐจ์› ์ถ•์†Œ ๊ธฐ๋ฒ•์ž…๋‹ˆ๋‹ค. ์ฃผ๋กœ ๊ณ ์ฐจ์› ๋ฐ์ดํ„ฐ์˜ ํด๋Ÿฌ์Šคํ„ฐ๋ง ๊ตฌ์กฐ๋ฅผ ์‹œ๊ฐ์ ์œผ๋กœ ํ‘œํ˜„ํ•˜๋Š” ๋ฐ ์‚ฌ์šฉ๋˜๋ฉฐ, ๋ฐ์ดํ„ฐ ํฌ์ธํŠธ ๊ฐ„์˜ ์ง€์—ญ์  ์œ ์‚ฌ์„ฑ์„ ๋ณด์กดํ•˜๋Š” ๋ฐ ๊ฐ•๋ ฅํ•œ ์„ฑ๋Šฅ์„ ๋ฐœํœ˜ํ•ฉ๋‹ˆ๋‹ค.t-SNE์˜ ํŠน์ง•t-SNE์˜ ํŠน์ง•์€ ์–ด๋– ํ•œ ์ ๋“ค์ด ์žˆ์„๊นŒ์š”?๋น„์„ ํ˜• ์ฐจ์› ์ถ•์†Œ: t-SNE๋Š” ๊ณ ์ฐจ์› ๋ฐ์ดํ„ฐ์˜ ๋น„์„ ํ˜• ๊ตฌ์กฐ๋ฅผ ์ €์ฐจ์›์—์„œ ๋ณด์กดํ•˜๋Š” ๋ฐ ํŠนํ™”๋˜์–ด ์žˆ์Šต๋‹ˆ๋‹ค. ์ด๋Š” ๋ฐ์ดํ„ฐ๊ฐ€ ์„ ํ˜•์ ์ธ ๊ตฌ์กฐ๋ฅผ ๊ฐ€์ง€์ง€ ์•Š๋Š” ๊ฒฝ์šฐ์—๋„ ์ ํ•ฉํ•œ ๊ฒฐ๊ณผ๋ฅผ ์ œ๊ณตํ•ฉ๋‹ˆ๋‹ค.์ง€์—ญ์  ์œ ์‚ฌ์„ฑ ๋ณด์กด: t-SNE๋Š” ๋ฐ์ดํ„ฐ ํฌ์ธํŠธ ๊ฐ„์˜ ์ง€์—ญ์  ์œ ์‚ฌ์„ฑ์„ ์œ ์ง€ํ•˜๋Š” ๋ฐ ์ค‘์ ..

๐Ÿ“‡ Machine Learning

[ML] Isomap (์•„์ด์†Œ๋งต)

์ด๋ฒˆ์—๋Š” Isomap์ด๋ผ๋Š” ๋จธ์‹ ๋Ÿฌ๋‹ ๊ธฐ๋ฒ•์— ๋ฐํ•˜์—ฌ ์•Œ์•„๋ณด๊ฒ ์Šต๋‹ˆ๋‹ค. ์•„์ด์†Œ๋งต(Isomap)์€ ๋น„์„ ํ˜• ์ฐจ์› ์ถ•์†Œ ๊ธฐ๋ฒ•์œผ๋กœ, ๊ณ ์ฐจ์› ๋ฐ์ดํ„ฐ์˜ ๊ธฐํ•˜ํ•™์  ๊ตฌ์กฐ๋ฅผ ๋ณด์กดํ•˜๋ฉด์„œ ์ €์ฐจ์›์œผ๋กœ ๋ณ€ํ™˜ํ•˜๋Š” ๋ฐฉ๋ฒ•์ž…๋‹ˆ๋‹ค.์ด ์•Œ๊ณ ๋ฆฌ์ฆ˜์€ ์ง€์˜ค๋ฐ์‹ ๊ฑฐ๋ฆฌ(Geodesic Distance)๋ฅผ ์‚ฌ์šฉํ•˜์—ฌ ๋ฐ์ดํ„ฐ ๊ฐ„์˜ ๊ฑฐ๋ฆฌ๋ฅผ ์ธก์ •ํ•˜๊ณ , ์ด๋ฅผ ๋ฐ”ํƒ•์œผ๋กœ ์ €์ฐจ์› ๊ณต๊ฐ„์—์„œ ๋ฐ์ดํ„ฐ์˜ ๊ตฌ์กฐ๋ฅผ ์‹œ๊ฐํ™”ํ•ฉ๋‹ˆ๋‹ค. Isomap์˜ ํŠน์ง• ๋น„์„ ํ˜• ์ฐจ์› ์ถ•์†Œ: Isomap์€ ๋ฐ์ดํ„ฐ์˜ ๋น„์„ ํ˜• ๊ตฌ์กฐ๋ฅผ ๋ณด์กดํ•˜๋ฉด์„œ ์ฐจ์›์„ ์ถ•์†Œํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค. ์ด๋Š” PCA์™€ ๊ฐ™์€ ์„ ํ˜• ์ฐจ์› ์ถ•์†Œ ๊ธฐ๋ฒ•์œผ๋กœ๋Š” ์–ด๋ ค์šด ๋ฐ์ดํ„ฐ์˜ ๋ณต์žกํ•œ ๊ตฌ์กฐ๋ฅผ ์ž˜ ์œ ์ง€ํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.์ง€์˜ค๋ฐ์‹ ๊ฑฐ๋ฆฌ ๊ธฐ๋ฐ˜: ๋ฐ์ดํ„ฐ ํฌ์ธํŠธ ๊ฐ„์˜ ์‹ค์ œ ๊ฑฐ๋ฆฌ(์ง€์˜ค๋ฐ์‹ ๊ฑฐ๋ฆฌ)๋ฅผ ์‚ฌ์šฉํ•˜์—ฌ ๋ฐ์ดํ„ฐ์˜ ๊ตฌ์กฐ์  ๊ด€๊ณ„๋ฅผ ๋ฐ˜์˜ํ•ฉ๋‹ˆ๋‹ค. ์ง€์˜ค๋ฐ์‹ ๊ฑฐ๋ฆฌ๋Š” ๋ฐ์ดํ„ฐ์˜..

๐Ÿ“‡ Machine Learning

[ML] Principal Component Analysis (PCA - ์ฃผ์„ฑ๋ถ„ ๋ถ„์„)

์ฃผ์„ฑ๋ถ„ ๋ถ„์„(Principal Component Analysis, PCA)์€ ๊ณ ์ฐจ์› ๋ฐ์ดํ„ฐ๋ฅผ ์ €์ฐจ์›์œผ๋กœ ๋ณ€ํ™˜ํ•˜์—ฌ ๋ฐ์ดํ„ฐ์˜ ์ฃผ์š” ๋ณ€๋™์„ฑ์„ ๋ณด์กดํ•˜๋Š” ์ฐจ์› ์ถ•์†Œ ๊ธฐ๋ฒ•์ž…๋‹ˆ๋‹ค. ์ด ๋ฐฉ๋ฒ•์€ ๋ฐ์ดํ„ฐ์˜ ๋ถ„์‚ฐ์„ ์ตœ๋Œ€ํ™”ํ•˜๋Š” ์ง๊ต ์ถ•์„ ์ฐพ์•„ ๋ฐ์ดํ„ฐ๋ฅผ ์ƒˆ๋กœ์šด ์ขŒํ‘œ๊ณ„๋กœ ๋ณ€ํ™˜ํ•จ์œผ๋กœ์จ ๋…ธ์ด์ฆˆ๋ฅผ ์ค„์ด๊ณ , ์‹œ๊ฐํ™” ๋ฐ ํ•ด์„์„ ์šฉ์ดํ•˜๊ฒŒ ํ•ฉ๋‹ˆ๋‹ค. ์ฃผ์„ฑ๋ถ„ ๋ถ„์„์€ ๋ฐ์ดํ„ฐ ์‹œ๊ฐํ™”, ๋…ธ์ด์ฆˆ ์ œ๊ฑฐ, ๋ฐ์ดํ„ฐ ์••์ถ• ๋“ฑ์˜ ๋ชฉ์ ์œผ๋กœ ๋„๋ฆฌ ์‚ฌ์šฉ๋ฉ๋‹ˆ๋‹ค.PCA(์ฃผ์„ฑ๋ถ„ ๋ถ„์„)์˜ ํŠน์ง•PCA์˜ ์ฃผ์š”ํ•œ ํŠน์ง•์€ ์–ด๋– ํ•œ ์ ์ด ์žˆ์„๊นŒ์š”? 1. ๋ถ„์‚ฐ ์ตœ๋Œ€ํ™”PCA๋Š” ๋ฐ์ดํ„ฐ์˜ ๋ถ„์‚ฐ์„ ์ตœ๋Œ€ํ™”ํ•˜๋Š” ๋ฐฉํ–ฅ์œผ๋กœ ์ƒˆ๋กœ์šด ์ถ•์„ ํƒ์ƒ‰ํ•ฉ๋‹ˆ๋‹ค. ๊ฐ€์žฅ ๋งŽ์€ ๋ณ€๋™์„ฑ์„ ์„ค๋ช…ํ•˜๋Š” ์ฃผ์„ฑ๋ถ„์„ ์ฐพ๋Š”๋‹ค๋Š” ์˜๋ฏธ์ž…๋‹ˆ๋‹ค.2. ์ง๊ต ์ถ•์ฃผ์„ฑ๋ถ„์€ ์„œ๋กœ ์ง๊ต(orthogonal)ํ•˜๋Š” ์ถ•์œผ๋กœ ๊ตฌ์„ฑ๋ฉ๋‹ˆ๋‹ค. ์ด๋กœ ์ธํ•ด ์ฃผ์„ฑ๋ถ„ ..

๐Ÿ“‡ Machine Learning

[ML] DBSCAN (Density-Based Spatial Clustering of Applications with Noise)

DBSCAN์€ ๋ฐ€๋„ ๊ธฐ๋ฐ˜์˜ ๊ตฐ์ง‘ํ™” ์•Œ๊ณ ๋ฆฌ์ฆ˜์œผ๋กœ, ๋ฐ์ดํ„ฐ ํฌ์ธํŠธ์˜ ๋ฐ€์ง‘๋œ ์˜์—ญ์„ ๊ตฐ์ง‘์œผ๋กœ ์‹๋ณ„ํ•˜๊ณ , ๋ฐ€๋„๊ฐ€ ๋‚ฎ์€ ๋ฐ์ดํ„ฐ ํฌ์ธํŠธ๋Š” ๋…ธ์ด์ฆˆ๋กœ ๊ฐ„์ฃผํ•˜๋Š” ๋ฐฉ์‹์ž…๋‹ˆ๋‹ค. DBSCAN์˜ ๋ชฉํ‘œ๋Š” ๋ฐ์ดํ„ฐ์˜ ๋ฐ€์ง‘ ์˜์—ญ์„ ์ฐพ์•„๋‚ด์–ด, ๊ตฐ์ง‘์˜ ํฌ๊ธฐ๋‚˜ ํ˜•ํƒœ์— ๊ตฌ์• ๋ฐ›์ง€ ์•Š๊ณ  ์œ ์—ฐํ•˜๊ฒŒ ๊ตฐ์ง‘ํ™”๋ฅผ ์ˆ˜ํ–‰ํ•˜๋Š” ๊ฒƒ์ž…๋‹ˆ๋‹ค.DBSCAN์˜ ํŠน์ง•๋ฐ€๋„ ๊ธฐ๋ฐ˜ ๊ตฐ์ง‘ํ™”: DBSCAN์€ ๋ฐ์ดํ„ฐ ํฌ์ธํŠธ์˜ ๋ฐ€๋„๋ฅผ ๊ธฐ์ค€์œผ๋กœ ๊ตฐ์ง‘์„ ํ˜•์„ฑํ•ฉ๋‹ˆ๋‹ค. ์ฆ‰, ์ผ์ • ๋ฐ€๋„ ์ด์ƒ์˜ ์˜์—ญ์„ ํ•˜๋‚˜์˜ ๊ตฐ์ง‘์œผ๋กœ ๋ฌถ์Šต๋‹ˆ๋‹ค.๋…ธ์ด์ฆˆ ์ฒ˜๋ฆฌ: ๋ฐ€๋„๊ฐ€ ๋‚ฎ์€ ๋ฐ์ดํ„ฐ ํฌ์ธํŠธ๋Š” ๋…ธ์ด์ฆˆ๋กœ ๊ฐ„์ฃผ๋˜๋ฉฐ, ๊ตฐ์ง‘์—์„œ ์ œ์™ธ๋ฉ๋‹ˆ๋‹ค. ์ด๋ฅผ ํ†ตํ•ด ๊ตฐ์ง‘ํ™” ๊ณผ์ •์—์„œ ๋…ธ์ด์ฆˆ๋‚˜ ์ด์ƒ์น˜๋ฅผ ํšจ๊ณผ์ ์œผ๋กœ ์ฒ˜๋ฆฌํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.์œ ์—ฐํ•œ ๊ตฐ์ง‘ ํ˜•์„ฑ: DBSCAN์€ ๊ตฐ์ง‘์˜ ํฌ๊ธฐ๋‚˜ ํ˜•ํƒœ์— ๊ตฌ์• ๋ฐ›์ง€ ์•Š๊ณ , ๋ฐ์ดํ„ฐ์˜ ๋ฐ€๋„์— ๋”ฐ๋ผ ์œ ์—ฐํ•˜๊ฒŒ ๊ตฐ์ง‘์„ ํ˜•..

๐Ÿ“‡ Machine Learning

[ML] Hierarchical Clustering (๊ณ„์ธต์  ๊ตฐ์ง‘ ๋ถ„์„)

Hierarchical Clustering (๊ณ„์ธต์  ๊ตฐ์ง‘ ๋ถ„์„)๋„ Unsupervised Learning (๋น„์ง€๋„ ํ•™์Šต) ๊ณ„์ธต์  ๊ตฐ์ง‘ ๋ถ„์„์€ ๋ฐ์ดํ„ฐ ํฌ์ธํŠธ๋“ค ๊ฐ„์˜ ์œ ์‚ฌ๋„๋ฅผ ๋ฐ”ํƒ•์œผ๋กœ ๊ณ„์ธต์ ์ธ ๊ตฐ์ง‘ ๊ตฌ์กฐ๋ฅผ ํ˜•์„ฑํ•˜๋Š” ๊ตฐ์ง‘ํ™” ๋ฐฉ๋ฒ•์ž…๋‹ˆ๋‹ค.์ด ๋ฐฉ๋ฒ•์€ ๋ฐ์ดํ„ฐ๋ฅผ ํŠธ๋ฆฌ ๊ตฌ์กฐ๋กœ ํ‘œํ˜„ํ•˜๋ฉฐ, ๋‹จ๊ณ„๋ณ„๋กœ ๊ตฐ์ง‘ํ™”๋ฅผ ์ง„ํ–‰ํ•จ์œผ๋กœ์จ ๋ฐ์ดํ„ฐ ๊ฐ„์˜ ๊ด€๊ณ„์™€ ๊ตฌ์กฐ๋ฅผ ์ดํ•ดํ•˜๋Š” ๋ฐ ๋„์›€์„ ์ค๋‹ˆ๋‹ค.Hierarchical Clustering (๊ณ„์ธต์  ๊ตฐ์ง‘ ๋ถ„์„)์˜ ์œ ํ˜•๊ทธ๋Ÿฌ๋ฉด, Hierarchical Clustering (๊ณ„์ธต์  ๊ตฐ์ง‘ ๋ถ„์„)์˜ ์œ ํ˜•์€ ์–ด๋– ํ•œ ๊ฒƒ์ด ์žˆ์„๊นŒ์š”? ํ•œ๋ฒˆ ์•Œ์•„๋ณด๊ฒ ์Šต๋‹ˆ๋‹ค. 1. ๋ณ‘ํ•ฉ์  ๊ตฐ์ง‘ํ™” (Agglomerative Clustering)๋ณ‘ํ•ฉ์  ๊ตฐ์ง‘ํ™”๋Š” ๊ฐ ๋ฐ์ดํ„ฐ๋ฅผ ํ•˜๋‚˜์˜ ๊ตฐ์ง‘์œผ๋กœ ์‹œ์ž‘ํ•˜์—ฌ, ๊ฐ€์žฅ ๊ฐ€๊นŒ์šด ๊ตฐ์ง‘๋“ค์„ ๋ฐ˜๋ณต์ ์œผ๋กœ ๋ณ‘ํ•ฉํ•ด..

Bigbread1129
'๐Ÿ“‡ Machine Learning' ์นดํ…Œ๊ณ ๋ฆฌ์˜ ๊ธ€ ๋ชฉ๋ก