{"id":59,"date":"2025-10-02T15:50:19","date_gmt":"2025-10-02T08:50:19","guid":{"rendered":"https:\/\/wise-pink-goat.210-211-125-205.cpanel.site\/wp\/?p=59"},"modified":"2025-10-02T15:50:19","modified_gmt":"2025-10-02T08:50:19","slug":"tong-hop-tat-ca-cac-thuat-toan-trong-machine-learning-phan-2","status":"publish","type":"post","link":"https:\/\/wise-pink-goat.210-211-125-205.cpanel.site\/wp\/tong-hop-tat-ca-cac-thuat-toan-trong-machine-learning-phan-2\/","title":{"rendered":"T\u1ed5ng h\u1ee3p t\u1ea5t c\u1ea3 c\u00e1c thu\u1eadt to\u00e1n trong Machine Learning (Ph\u1ea7n 2)"},"content":{"rendered":"\n

1. Bayes Theorem (\u0110\u1ecbnh l\u00fd Bayes)<\/h3>\n\n\n\n
\\begin{aligned}\nP(A\\mid B) &= \\frac{P(B\\mid A)P(A)}{P(B)}\n\\end{aligned}\n\\\\\n\\begin{align*}\n\\\\\n&P(A\\mid B): \\text{X\u00e1c su\u1ea5t x\u1ea3y ra A bi\u1ebft B \u0111\u00e3 x\u1ea3y ra r\u1ed3i}\\\\\n&P(B\\mid A): \\text{X\u00e1c su\u1ea5t x\u1ea3y ra B bi\u1ebft A \u0111\u00e3 x\u1ea3y ra r\u1ed3i}\\\\\n&P(A): \\text{X\u00e1c su\u1ea5t x\u1ea3y ra A}\\\\\n&P(B): \\text{X\u00e1c su\u1ea5t x\u1ea3y ra B}\\\\\n\\end{align*}\n<\/pre><\/div>\n\n\n\n
C\u00e1c bi\u1ebfn th\u1ec3 c\u1ee7a Naive Bayes:<\/p>\n\n\n\n
    \n
  • Multinomial Naive Bayes: Input \u0111\u1ea7u v\u00e0o l\u00e0 c\u00e1c gi\u00e1 tr\u1ecb r\u1eddi r\u1ea1c b\u1eb1ng s\u1ed1 l\u1ea7n c\u00e1c gi\u00e1 tr\u1ecb xu\u1ea5t hi\u1ec7n trong \u0111o\u1ea1n text.<\/li>\n\n\n\n
  • Bernoulli Naive Bayes: C\u00e1c t\u1eeb \u0111\u01b0\u1ee3c x\u00e1c \u0111\u1ecbnh b\u1eb1ng vi\u1ec7c ch\u00fang xu\u1ea5t hi\u1ec7n hay l\u00e0 kh\u00f4ng.<\/li>\n\n\n\n
  • Gaussian Naive Bayes: Input \u0111\u1ea7u v\u00e0o l\u00e0 bi\u1ebfn li\u00ean t\u1ee5c + gi\u1ea3 thuy\u1ebft ph\u00e2n ph\u1ed1i chu\u1ea9n.<\/li>\n<\/ul>\n\n\n\n
    2. Decision Tree (C\u00e2y quy\u1ebft \u0111\u1ecbnh)<\/h3>\n\n\n\n
    Thu\u1eadt to\u00e1n \u0111\u01b0\u1ee3c h\u00ecnh th\u00e0nh t\u1eeb c\u00e1c Decision Node<\/code> v\u00e0 Leaf Node<\/code>, v\u00e0 1 node \u0111\u1eb7c bi\u1ec7t l\u00e0 Root Node<\/code>.<\/p>\n\n\n\n
    Vi\u1ec7c x\u00e2y d\u1ef1ng thu\u1eadt to\u00e1n d\u1ef1a v\u00e0o nhi\u1ec1u y\u1ebfu t\u1ed1:<\/p>\n\n\n\n
      \n
    • \u0110\u1ed9 s\u00e2u c\u1ee7a tree.<\/li>\n\n\n\n
    • \u0110\u1ed9 l\u1ec7ch c\u1ee7a Leaf Node<\/code> khi \u0111\u01b0a ra quy\u1ebft \u0111\u1ecbnh t\u1ed1i thi\u1ec3u l\u00e0 bao nhi\u00eau, \u0111\u1ed9 l\u1ec7ch c\u00e0ng l\u1edbn c\u00e0ng t\u1ed1t \u0111\u1ec3 ch\u1eafc ch\u1eafn v\u1ec1 d\u1ef1 \u0111o\u00e1n.<\/li>\n\n\n\n
    • M\u1ed7i Decision Node<\/code> n\u00ean ch\u1ecdn nh\u1eefng feature n\u00e0o \u0111\u1ec3 c\u00f3 th\u1ec3 \u0111\u1eb7t c\u00e2u h\u1ecfi, ph\u00e2n nh\u00e1nh d\u1eef li\u1ec7u. N\u00ean ch\u1ecdn c\u00e1c feature sao cho ph\u00e2n t\u00e1ch d\u1eef li\u1ec7u c\u00e0ng l\u1ec7ch c\u00e0ng t\u1ed1t. C\u00e1ch ph\u1ed5 bi\u1ebfn \u0111\u1ec3 l\u1ef1a ch\u1ecdn feature l\u00e0 s\u1eed d\u1ee5ng Gini Impurity.<\/li>\n<\/ul>\n\n\n\n
      \\begin{aligned}\n\\text{Gini Impurity}\\\\\n\\\\\nGini = 1 - \\sum_{i=1}^C(P_i)^2\n\\end{aligned}\n\\\\\n\\begin{align*}\n\\\\\n&C: \\text{T\u1ed5ng s\u1ed1 l\u01b0\u1ee3ng class}\\\\\n&P_i: \\text{X\u00e1c su\u1ea5t 1 ph\u1ea7n t\u1eed thu\u1ed9c v\u1ec1 class i}\n\\end{align*}\n<\/pre><\/div>\n\n\n\n
        \n
      • N\u1ebfu 1 feature gi\u00fap ph\u00e2n chia node c\u00e0ng l\u1ec7ch th\u00ec Gini c\u00e0ng th\u1ea5p v\u00e0 ng\u01b0\u1ee3c l\u1ea1i. Gini th\u1ea5p nh\u1ea5t l\u00e0 = 0, cao nh\u1ea5t l\u00e0 = 0.5, c\u00e0ng th\u1ea5p c\u00e0ng t\u1ed1t, c\u00e0ng cao c\u00e0ng t\u1ed3i.<\/li>\n<\/ul>\n\n\n\n
        V\u00ed d\u1ee5 v\u1ec1 vi\u1ec7c \u00e1p d\u1ee5ng c\u00f4ng th\u1ee9c, ta c\u00f3 5 ng\u01b0\u1eddi chia l\u00e0m 2 class c\u00f3 cho vay v\u00e0 kh\u00f4ng cho vay. C\u00f3 3 ng\u01b0\u1eddi l\u00e0 kh\u00f4ng cho vay v\u00e0 c\u00f3 2 ng\u01b0\u1eddi l\u00e0 kh\u00f4ng cho vay. X\u00e1c su\u1ea5t cho vay l\u00e0 40%, x\u00e1c su\u1ea5t kh\u00f4ng cho vay l\u00e0 60%. \u00c1p d\u1ee5ng c\u00f4ng th\u1ee9c ta c\u00f3:<\/p>\n\n\n\n
        \\begin{aligned}\n&1- (0.4^2 + 0.6^2) = 0.48\n\\\\\n&\\text{\u0110\u00e2y l\u00e0 1 gi\u00e1 tr\u1ecb kh\u00e1 t\u1ed3i do n\u00f3 g\u1ea7n 0.5 => ph\u00e2n chia \u0111\u1ed3ng \u0111\u1ec1u}\n\\end{aligned}<\/pre><\/div>\n\n\n\n
        Ngo\u00e0i s\u1eed d\u1ee5ng Gini Impurity, ta c\u00f2n c\u00f3 th\u1ec3 s\u1eed d\u1ee5ng Entropy \u0111\u1ec3 quy\u1ebft \u0111\u1ecbnh xem \u1edf m\u1ed7i Decision Node<\/code>, feature n\u00e0o l\u00e0 t\u1ed1t nh\u1ea5t:<\/p>\n\n\n\n
        \\begin{aligned}\nEntropy = \\sum_{i=1}^C - P_i * \\log_2(P_i)\n\\end{aligned}<\/pre><\/div>\n\n\n\n
        Ngo\u00e0i \u01b0u \u0111i\u1ec3m d\u1ec5 nh\u00ecn v\u00e0 tr\u1ef1c quan, Decision Tree c\u00f2n 1 s\u1ed1 nh\u01b0\u1ee3c \u0111i\u1ec3m sau:<\/p>\n\n\n\n
          \n
        • R\u1ea5t d\u1ec5 b\u1ecb overfitting khi c\u00e2y qu\u00e1 s\u00e2u, qu\u00e1 nhi\u1ec1u level.<\/li>\n\n\n\n
        • Ch\u1ec9 1 thay \u0111\u1ed5i nh\u1ecf c\u1ee7a data c\u0169ng d\u1eabn \u0111\u1ebfn thay \u0111\u1ed5i c\u1ea3 c\u1ea5u tr\u00fac tree.<\/li>\n\n\n\n
        • \u00cdt \u0111\u01b0\u1ee3c s\u1eed d\u1ee5ng trong b\u00e0i to\u00e1n Regression v\u00ec r\u1ea5t kh\u00f3 \u0111\u1ec3 t\u00ecm \u0111\u01b0\u1ee3c \u0111i\u1ec3m c\u00e2n b\u1eb1ng gi\u1eefa overfitting v\u00e0 underfitting (d\u1eef li\u1ec7u qu\u00e1 ph\u1ee9c t\u1ea1p c\u00f2n m\u00f4 h\u00ecnh qu\u00e1 \u0111\u01a1n gi\u1ea3n n\u00ean kh\u00f4ng \u0111\u1ee7 kh\u1ea3 n\u0103ng t\u1ed5ng qu\u00e1t h\u00f3a xu h\u01b0\u1edbng d\u1eef li\u1ec7u).<\/li>\n<\/ul>\n\n\n\n
          D\u00f9ng trong b\u00e0i to\u00e1n n\u00e0o th\u00ec Decision Tree d\u00f9ng 1 m\u00ecnh s\u1ebd c\u00f3 \u0111\u1ed9 ch\u00ednh x\u00e1c kh\u00f4ng cao, do \u0111\u00f3 s\u1ebd k\u1ebft h\u1ee3p nhi\u1ec1u Decision Tree l\u1ea1i v\u1edbi nhau \u0111\u1ec3 t\u0103ng \u0111\u1ed9 ch\u00ednh x\u00e1c.<\/p>\n\n\n\n
          3. Random Forest (r\u1eebng ng\u1eabu nhi\u00ean)<\/h3>\n\n\n\n
          Random Forest s\u1ebd t\u1ed5ng h\u1ee3p c\u00e1c d\u1ef1 \u0111o\u00e1n c\u1ee7a nhi\u1ec1u Decision Tree \u0111\u1ec3 \u0111\u01b0a ra d\u1ef1 \u0111o\u00e1n cu\u1ed1i c\u00f9ng.<\/p>\n\n\n\n
            \n
          • V\u1edbi Classification s\u1ebd d\u00f9ng Majority Vote, ngh\u0129a l\u00e0 class n\u00e0o \u0111\u01b0\u1ee3c \u0111a s\u1ed1 vote th\u00ec s\u1ebd \u0111\u01b0\u1ee3c ch\u1ecdn l\u00e0m Final Prediction.<\/li>\n\n\n\n
          • V\u1edbi Regression s\u1ebd d\u00f9ng Averaging, ngh\u0129a l\u00e0 l\u1ea5y gi\u00e1 tr\u1ecb trung b\u00ecnh c\u1ee7a c\u00e1c d\u1ef1 \u0111o\u00e1n c\u1ee7a Decision Tree \u0111\u1ec3 t\u00ecm Final Prediction.<\/li>\n<\/ul>\n\n\n\n
            \u0110\u1ec3 vi\u1ec7c k\u1ebft h\u1ee3p Decision Tree c\u00f3 \u00fd ngh\u0129a th\u00ec m\u1ed7i ch\u00fang ph\u1ea3i \u0111\u01b0\u1ee3c hu\u1ea5n luy\u1ec7n tr\u00ean 1 b\u1ed9 d\u1eef li\u1ec7u kh\u00e1c nhau \u0111\u1ec3 tr\u00e1nh vi\u1ec7c t\u1ea5t c\u1ea3 h\u1ecdc v\u00e0 \u0111\u01b0a d\u1ef1 \u0111o\u00e1n gi\u1ed1ng h\u1ec7t nhau. \u0110\u1ec3 c\u00f3 nhi\u1ec1u b\u1ed9 data kh\u00e1c nhau ta c\u1ea7n \u00e1p d\u1ee5ng Boostrapping \u0111\u1ec3 l\u1ea5y m\u1eabu c\u00f3 ho\u00e0n l\u1ea1i. T\u1eeb b\u1ed9 data ban \u0111\u1ea7u, l\u1ea5y m\u1eabu c\u00f3 ho\u00e0n l\u1ea1i nhi\u1ec1u l\u1ea7n \u0111\u1ec3 t\u1ea1o b\u1ed9 d\u1eef li\u1ec7u con r\u1ed3i \u0111em \u0111i hu\u1ea5n luy\u1ec7n nhi\u1ec1u Decision Tree.<\/p>\n\n\n\n
            \"\"<\/figure>\n\n\n\n
            M\u1ed9t l\u01b0u \u00fd l\u00e0 m\u1ed7i Decision Node ch\u1ec9 \u0111\u01b0\u1ee3c ch\u1ecdn feature t\u1ed1t nh\u1ea5t trong 1 t\u1eadp con c\u1ee7a c\u00e1c feature \u0111\u1ec3 t\u0103ng \u0111\u1ed9 \u0111a d\u1ea1ng d\u1ef1 \u0111o\u00e1n.<\/p>\n\n\n\n
            Random Forest c\u00f3 t\u00ednh gi\u1ea3i th\u00edch <\/mark>th\u1ea5p h\u01a1n. T\u00ednh gi\u1ea3i th\u00edch l\u00e0 sau khi d\u1ef1 \u0111o\u00e1n cu\u1ed1i c\u00f9ng c\u00f3 th\u1ec3 quay l\u1ea1i gi\u1ea3i th\u00edch t\u1ea1i sao m\u00f4 h\u00ecnh l\u1ea1i c\u00f3 d\u1ef1 \u0111o\u00e1n \u0111\u00f3. Random Forest c\u0169ng t\u1ed1n nhi\u1ec1u th\u1eddi gian v\u00e0 t\u00e0i nguy\u00ean h\u01a1n do ta ph\u1ea3i hu\u1ea5n luy\u1ec7n nhi\u1ec1u Decision Tree c\u00f9ng l\u00fac.<\/p>\n\n\n\n
            4. Support Vector Machine<\/h3>\n\n\n\n
            M\u1ee5c ti\u00eau c\u1ee7a SVM trong b\u00e0i to\u00e1n classification l\u00e0 t\u00ecm ra 1 \u0111\u01b0\u1eddng ph\u00e2n c\u00e1ch t\u1ed1t nh\u1ea5t \u0111\u1ec3 ph\u00e2n chia c\u00e1c \u0111i\u1ec3m d\u1eef li\u1ec7u c\u1ee7a c\u00e1c class kh\u00e1c nhau trong kh\u00f4ng gian \u0111a chi\u1ec1u.<\/p>\n\n\n\n
            \u0110\u1ec3 l\u00e0 1 \u0111\u01b0\u1eddng ph\u00e2n c\u00e1ch t\u1ed1t nh\u1ea5t th\u00ec c\u1ea7n th\u1ecfa m\u00e3n 2 \u0111i\u1ec1u ki\u1ec7n:<\/p>\n\n\n\n
              \n
            • C\u00e1ch \u0111\u1ec1u c\u00e1c \u0111i\u1ec3m g\u1ea7n nh\u1ea5t c\u1ee7a 2 class g\u1ecdi l\u00e0 Margin.<\/li>\n\n\n\n
            • Margin \u0111\u01b0\u1ee3c t\u1ea1o ra l\u00e0 t\u1ed1i \u0111a.<\/li>\n<\/ul>\n\n\n\n
              C\u00e1c \u0111i\u1ec3m d\u1eef li\u1ec7u g\u1ea7n \u0111\u01b0\u1eddng ph\u00e2n c\u00e1ch t\u1ed1i \u01b0u nh\u1ea5t c\u1ee7a c\u00e1c class g\u1ecdi l\u00e0 Support Vector v\u00ec n\u00f3 s\u1ebd quy\u1ebft \u0111\u1ecbnh c\u00f4ng th\u1ee9c t\u00ednh \u0111\u01b0\u1eddng ph\u00e2n c\u00e1ch t\u1ed1i \u01b0u. C\u00e1c \u0111\u01b0\u1eddng ph\u00e2n c\u00e1ch g\u1ecdi l\u00e0 Hyperplane (si\u00eau ph\u1eb3ng), \u0111\u01b0\u1eddng ph\u00e2n c\u00e1ch t\u1ed1i \u01b0u \u0111\u01b0\u1ee3c g\u1ecdi l\u00e0 Optimal Hyperplane (si\u00eau ph\u1eb3ng t\u1ed1i \u01b0u).<\/p>\n\n\n\n
              Trong c\u00e1c b\u1ed9 d\u1eef li\u1ec7u th\u1ef1c t\u1ebf s\u1ebd c\u00f3 2 c\u00e1ch th\u1ee9c m\u00e0 SVM s\u1ebd ph\u1ea3i \u0111\u1ed1i m\u1eb7t:<\/p>\n\n\n\n
                \n
              • Kh\u00f4ng ph\u1ea3i l\u00fac n\u00e0o c\u0169ng t\u00ecm \u0111\u01b0\u1ee3c \u0111\u01b0\u1eddng ph\u00e2n c\u00e1ch t\u1ed1i \u01b0u \u0111\u1ec3 ph\u00e2n chia ho\u00e0n h\u1ea3o c\u00e1c \u0111i\u1ec3m d\u1eef li\u1ec7u m\u00e0 kh\u00f4ng sai s\u00f3t ch\u00fat n\u00e0o. \u0110\u1ec3 kh\u1eafc ph\u1ee5c ta s\u1ebd ch\u1ecdn 1 gi\u00e1 tr\u1ecb l\u00e0 Hyperparameter g\u1ecdi l\u00e0 C. C l\u00e0 ng\u01b0\u1ee1ng m\u00e0 SVM s\u1ebd ch\u1ea5p nh\u1eadn l\u00e0 s\u1ebd sai mi\u1ec5n l\u00e0 d\u01b0\u1edbi C l\u00e0 \u0111\u01b0\u1ee3c.<\/li>\n\n\n\n
              • Kh\u00f4ng ph\u1ea3i l\u00fac n\u00e0o d\u1eef li\u1ec7u c\u0169ng \u0111\u01b0\u1ee3c ph\u00e2n chia 1 c\u00e1ch tuy\u1ebfn t\u00ednh. Ngh\u0129a l\u00e0 trong 2D kh\u00f4ng ph\u1ea3i l\u00fac n\u00e0o c\u0169ng k\u1ebb \u0111\u01b0\u1ee3c 1 \u0111\u01b0\u1eddng th\u1eb3ng \u0111\u1ec3 c\u00f3 th\u1ec3 ph\u00e2n chia d\u1eef li\u1ec7u v\u1ec1 2 ph\u00eda c\u1ee7a \u0111\u01b0\u1eddng th\u1eb3ng. L\u00fac n\u00e0y s\u1ebd \u00e1p d\u1ee5ng k\u1ef9 thu\u1eadt g\u1ecdi l\u00e0 Kernel Trick \u0111\u1ec3 \u0111\u01b0a t\u1eeb 2D l\u00ean nhi\u1ec1u chi\u1ec1u h\u01a1n \u0111\u1ec3 d\u1ec5 ph\u00e2n chia tuy\u1ebfn t\u00ednh.<\/li>\n<\/ul>\n\n\n\n
                SVM c\u00f3 hi\u1ec7u n\u0103ng cao, l\u00e0m vi\u1ec7c t\u1ed1t v\u1edbi data nhi\u1ec1u chi\u1ec1u. Nh\u01b0ng n\u00f3 s\u1ebd ch\u1eadm ch\u1ea1p khi \u00e1p d\u1ee5ng l\u00ean data c\u00f3 qu\u00e1 nhi\u1ec1u data point.<\/p>\n","protected":false},"excerpt":{"rendered":"
                1. Bayes Theorem (\u0110\u1ecbnh l\u00fd Bayes) C\u00e1c bi\u1ebfn th\u1ec3 c\u1ee7a Naive Bayes: 2. Decision Tree (C\u00e2y quy\u1ebft \u0111\u1ecbnh) Thu\u1eadt to\u00e1n \u0111\u01b0\u1ee3c h\u00ecnh th\u00e0nh t\u1eeb c\u00e1c Decision Node v\u00e0 Leaf Node, v\u00e0 1 node \u0111\u1eb7c bi\u1ec7t l\u00e0 Root Node. Vi\u1ec7c x\u00e2y d\u1ef1ng thu\u1eadt to\u00e1n d\u1ef1a v\u00e0o nhi\u1ec1u y\u1ebfu t\u1ed1: V\u00ed d\u1ee5 v\u1ec1 vi\u1ec7c \u00e1p d\u1ee5ng c\u00f4ng […]<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"pagelayer_contact_templates":[],"_pagelayer_content":"","footnotes":""},"categories":[1],"tags":[4,3],"class_list":["post-59","post","type-post","status-publish","format-standard","hentry","category-machine-learning","tag-algorithm","tag-machine-learning","entry"],"_links":{"self":[{"href":"https:\/\/wise-pink-goat.210-211-125-205.cpanel.site\/wp\/wp-json\/wp\/v2\/posts\/59","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/wise-pink-goat.210-211-125-205.cpanel.site\/wp\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/wise-pink-goat.210-211-125-205.cpanel.site\/wp\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/wise-pink-goat.210-211-125-205.cpanel.site\/wp\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/wise-pink-goat.210-211-125-205.cpanel.site\/wp\/wp-json\/wp\/v2\/comments?post=59"}],"version-history":[{"count":16,"href":"https:\/\/wise-pink-goat.210-211-125-205.cpanel.site\/wp\/wp-json\/wp\/v2\/posts\/59\/revisions"}],"predecessor-version":[{"id":79,"href":"https:\/\/wise-pink-goat.210-211-125-205.cpanel.site\/wp\/wp-json\/wp\/v2\/posts\/59\/revisions\/79"}],"wp:attachment":[{"href":"https:\/\/wise-pink-goat.210-211-125-205.cpanel.site\/wp\/wp-json\/wp\/v2\/media?parent=59"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/wise-pink-goat.210-211-125-205.cpanel.site\/wp\/wp-json\/wp\/v2\/categories?post=59"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/wise-pink-goat.210-211-125-205.cpanel.site\/wp\/wp-json\/wp\/v2\/tags?post=59"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}