DevCoach 185: Machine Learning | Supervised Learning

Sesi ini akan membahas Sentimen Analisis sebagai bagian dari Natural Language Processing (NLP), mulai dari konsep dasar hingga implementasi praktis. Peserta akan diperkenalkan pada teknik NLP yang digunakan untuk memahami dan mengolah teks, termasuk tokenisasi, stemming, dan word embeddings. Selanjutnya, akan dijelaskan pendekatan pemodelan menggunakan Machine Learning tradisional seperti Naïve Bayes dan SVM, serta eksplorasi model Deep Learning seperti LSTM dan Transformer untuk meningkatkan akurasi analisis sentimen. Melalui sesi ini, peserta akan memperoleh pemahaman tentang bagaimana membangun pipeline Sentimen Analisis yang efektif dan aplikatif di berbagai industri.

Transcript

1. Machine Learning Muhammad Fikry Rizal Curriculum Developer Machine Learning Supervised Learning
2. Hi, I’m M. Fikry Rizal 👋 Latest Work Experiences: • AI/ML Curriculum Developer, Dicoding 2024 - present • Data Engineer Intern, Torche Education 2022 - 2023 Education: • UIN Syarif Hidayatullah Jakarta 2020 - 2024 Bachelor Degree, Physics • Bangkit Academy 2023 2023 Machine Learning About Me Muhammad Fikry Rizal https://github.com /mfikryrz Machine Learning
3. Machine Learning Overview 1. Hi, Machine Learning! 2. Machine Learning Workflow 3. Supervised Learning: Klasifikasi 4. Supervised Learning: Regresi 5. Unsupervised Learning - Clustering 6. Teknik Feature Engineering 7. Overfitting dan Underfitting 8. Optimasi Model Machine Learning
4. Hi, Machine Learning! 1. Hi, Machine Learning! 2. Machine Learning Workflow 3. Supervised Learning: Classification 4. Supervised Learning: Regression Machine Learning
5. “A field of study that gives computers the ability to learn without being explicitly programmed.” Arthur Samuel Machine Learning
6. Categories of Machine Learning Machine Learning
7. Practical Use Cases of Machine Learning Machine Learning
8. Machine Learning Tools Machine Learning
9. Machine Learning Workflow Machine Learning
10. Data Collecting Machine Learning
11. Understanding Data: Sources and Formats • UC Irvine Machine Learning Repository • Kaggle Dataset • Google Dataset Search Engine • TensorFlow Dataset • Satu Data Indonesia • Menggunakan Dataset dari Sumber Terpilih Machine Learning • CSV Comma-Separated Values) • Excel Files • JSON JavaScript Object Notation) • HTML • SQL Database
12. Refining Raw Data: The Art of Data Cleaning Machine Learning Machine Learning
13. Exploratory & Explanatory Data Analysis • Exploratory Data Analysis EDA ◦ Understand structur, characteristics, and pattern in the data • Explanatory Data Analysis ExDA ◦ Communicating findings or insights that have been obtained to a broader audience. Machine Learning
14. Exploratory Machine Learning Explanatory
15. Refining Raw Data: The Art of Data Cleaning • Identifying and Handling Missing Values Machine Learning
16. Refining Raw Data: The Art of Data Cleaning Machine Learning Feature Scaling • Normalization ◦ Features with Different Scales. ◦ Distance-Based Models. ◦ Data Not Normally Distributed. • Standardization ◦ Normally Distributed Data. ◦ Regression-Based Models. ◦ Data with Different Scales.
17. Data Splitting Machine Learning • Training Set ◦ A subset of data used to train the model. ◦ Common Percentage: Typically 6080% of the total dataset. • Validation Set ◦ A subset of data used for validation during the training process. ◦ Common Percentage: Typically 1020% of the total dataset. • Test Set ◦ A subset of data used for final testing after the model has been trained and tuned. ◦ Common Percentage: Typically 1020% of the total dataset.
18. Machine Learning Crafting Powerful Machine Learning Models
19. Deployment & Monitoring Machine Learning
20. Quiz #1 DevCoach 185 Dalam tahap preprocessing data, Anda menemukan bahwa beberapa fitur memiliki missing values. Pilihan berikut mana yang paling tepat untuk menangani missing values dalam konteks model machine learning? Machine Learning a). Menghapus fitur tersebut b). Mengisinya dengan mean atau median dari fitur tersebut
21. Supervised Learning: Klasifikasi Machine Learning
22. Classification Algorithms Machine Learning
23. Algorithm: Decision Tree Advantages • Can Handle Both Categorical and Numerical Data • No Need for Feature Scaling • Flexible and Customizable Disadvantages • Can Be Too Flexible Prone to Overfitting) • Sensitive to Noise • Can Grow Too Large Complex Trees) How It Works • Step 1 Initial Data Splitting • Step 2 Feature Selection and Data Partitioning • Step 3 Branch and Node Formation • Step 4 Creation of Leaf Nodes • Step 5 Using the Model for Prediction Machine Learning
24. Algorithm: Random Forest Advantages • High Accuracy • Robust Against Overfitting • Ability to Handle Imbalanced Data • Handles Missing Data • Identifies Important Features Disadvantages • High Memory Requirements • Low Interpretability • Slow Prediction Speed • Less Effective on Small Datasets • Long Training Time Machine Learning
25. Model Evaluation Machine Learning
26. Model Evaluation Misalkan kita memiliki hasil prediksi model untuk sebuah dataset dengan 100 email yang diklasifikasikan sebagai spam atau bukan spam (ham). Machine Learning
27. Model Evaluation Machine Learning
28. Hands-on Coding: Let's Build It Machine Learning
29. Quiz #2 DevCoach 185 Seorang pasien menjalani tes untuk mendeteksi kanker. Hasil tes menunjukkan bahwa pasien tidak memiliki kanker, tetapi sebenarnya pasien tersebut memiliki kanker. Apa jenis kesalahan dalam klasifikasi ini? a) True Positive b) True Negative c) False Positive d) False Negative Machine Learning
30. Supervised Learning: Regression Machine Learning
31. Machine Learning Regression Algorithm
32. Algorithm: Linear Regression Machine Learning
33. Model Evaluation Machine Learning
34. Hands-on Coding: Let's Build It Machine Learning
35. Quiz #3 DevCoach 185 Jika model menghasilkan hasil dalam bentuk persentase probabilitas, misalnya 85% kemungkinan hujan, apakah model tersebut lebih cenderung merupakan model klasifikasi atau regresi? Machine Learning
36. Thank You Machine Learning
37. Feedback! dicoding.id/devcoachfeedback Machine Learning