Machine Learning and Data Science Blueprints for Finance

Spis treści książki

• Preface
  • Who This Book Is For
  • How This Book Is Organized
    • Part I: The Framework
    • Part II: Supervised Learning
    • Part III: Unsupervised Learning
    • Part IV: Reinforcement Learning and Natural Language Processing
  • Conventions Used in This Book
  • Using Code Presented in the Book
  • Python Libraries
  • OReilly Online Learning
  • How to Contact Us
  • Acknowledgments
    • Special Thanks from Hariom
    • Special Thanks from Sahil
    • Special Thanks from Brad
• I. The Framework
• 1. Machine Learning in Finance: The Landscape
  • Current and Future Machine Learning Applications in Finance
    • Algorithmic Trading
    • Portfolio Management and Robo-Advisors
    • Fraud Detection
    • Loans/Credit Card/Insurance Underwriting
    • Automation and Chatbots
    • Risk Management
    • Asset Price Prediction
    • Derivative Pricing
    • Sentiment Analysis
    • Trade Settlement
    • Money Laundering
  • Machine Learning, Deep Learning, Artificial Intelligence, and Data Science
  • Machine Learning Types
    • Supervised
      • Classification
      • Regression
    • Unsupervised
      • Dimensionality reduction
      • Clustering
    • Reinforcement Learning
  • Natural Language Processing
  • Chapter Summary
• 2. Developing a Machine Learning Model in Python
  • Why Python?
  • Python Packages for Machine Learning
    • Python and Package Installation
  • Steps for Model Development in Python Ecosystem
    • Model Development Blueprint
      • 1. Problem definition
      • 2. Loading the data and packages
        • 2.1. Load libraries
        • 2.2. Load data
      • 3. Exploratory data analysis
        • 3.1. Descriptive statistics
        • 3.2. Data visualization
      • 4. Data preparation
        • 4.1. Data cleaning
        • 4.2. Feature selection
        • 4.3. Data transformation
      • 5. Evaluate models
        • 5.1. Training and test split
        • 5.2. Identify evaluation metrics
        • 5.3. Compare models and algorithms
      • 6. Model tuning
      • 7. Finalize the model
        • 7.1. Performance on the test set
        • 7.2. Model/variable intuition
        • 7.3. Save/deploy
  • Chapter Summary
• 3. Artificial Neural Networks
  • ANNs: Architecture, Training, and Hyperparameters
    • Architecture
      • Neurons
      • Layers
        • Input layer
        • Hidden layers
        • Output layer
      • Neuron weights
    • Training
      • Forward propagation
      • Backpropagation
    • Hyperparameters
      • Number of hidden layers and nodes
      • Learning rate
      • Activation functions
      • Cost functions
      • Optimizers
      • Epoch
      • Batch size
  • Creating an Artificial Neural Network Model in Python
    • Installing Keras and Machine Learning Packages
      • Importing the packages
      • Loading data
      • Model constructiondefining the neural network architecture
      • Compiling the model
      • Fitting the model
      • Evaluating the model
    • Running an ANN Model Faster: GPU and Cloud Services
      • GPU
      • Cloud services such as Kaggle and Google Colab
  • Chapter Summary
• II. Supervised Learning
• 4. Supervised Learning: Models and Concepts
  • Supervised Learning Models: An Overview
    • Linear Regression (Ordinary Least Squares)
      • Implementation in Python
      • Training a model
      • Grid search
      • Advantages and disadvantages
    • Regularized Regression
    • Logistic Regression
      • Hyperparameters
      • Advantages and disadvantages
    • Support Vector Machine
      • Hyperparameters
      • Advantages and disadvantages
    • K-Nearest Neighbors
      • Hyperparameters
      • Advantages and disadvantages
    • Linear Discriminant Analysis
      • Implementation in Python and hyperparameters
      • Advantages and disadvantages
    • Classification and Regression Trees
      • Representation
      • Learning a CART model
      • Stopping criterion
      • Pruning the tree
      • Implementation in Python
      • Hyperparameters
      • Advantages and disadvantages
    • Ensemble Models
      • Random forest
      • Implementation in Python
      • Hyperparameters
      • Advantages and disadvantages
      • Extra trees
        • Implementation in Python
      • Adaptive Boosting (AdaBoost)
        • Implementation in Python
        • Hyperparameters
        • Advantages and disadvantages
      • Gradient boosting method
        • Implementation in Python and hyperparameters
        • Advantages and disadvantages
    • ANN-Based Models
      • ANN using sklearn
      • Hyperparameters
      • Deep neural network
      • Advantages and disadvantages
      • Using ANNs for supervised learning in finance
  • Model Performance
    • Overfitting and Underfitting
    • Cross Validation
    • Evaluation Metrics
      • Mean absolute error
      • Mean squared error
      • R metric
      • Adjusted R metric
      • Selecting an evaluation metric for supervised regression
      • Classification
      • Accuracy
      • Precision
      • Recall
      • Area under ROC curve
      • Confusion matrix
      • Selecting an evaluation metric for supervised classification
  • Model Selection
    • Factors for Model Selection
    • Model Trade-off
  • Chapter Summary
• 5. Supervised Learning: Regression (Including Time Series Models)
  • Time Series Models
    • Time Series Breakdown
    • Autocorrelation and Stationarity
      • Autocorrelation
      • Stationarity
      • Differencing
    • Traditional Time Series Models (Including the ARIMA Model)
      • ARIMA
    • Deep Learning Approach to Time Series Modeling
      • RNNs
      • Long short-term memory
    • Modifying Time Series Data for Supervised Learning Models
  • Case Study 1: Stock Price Prediction
    • Blueprint for Using Supervised Learning Models to Predict a Stock Price
      • 1. Problem definition
      • 2. Getting startedloading the data and Python packages
        • 2.1. Loading the Python packages
        • 2.2. Loading the data
      • 3. Exploratory data analysis
        • 3.1. Descriptive statistics
        • 3.2. Data visualization
        • 3.3. Time series analysis
      • 4. Data preparation
      • 5. Evaluate models
        • 5.1. Train-test split and evaluation metrics
        • 5.2. Test options and evaluation metrics
        • 5.3. Compare models and algorithms
        • 5.3.1. Machine learning models from Scikit-learn
        • 5.3.2. Time seriesbased models: ARIMA and LSTM
      • 6. Model tuning and grid search
      • 7. Finalize the model
        • 7.1. Results on the test dataset
      • Conclusion
  • Case Study 2: Derivative Pricing
    • Blueprint for Developing a Machine Learning Model for Derivative Pricing
      • 1. Problem definition
      • 2. Getting startedloading the data and Python packages
        • 2.1. Loading the Python packages
        • 2.2. Defining functions and parameters
        • Volatility and option pricing functions
        • 2.3. Data generation
      • 3. Exploratory data analysis
        • 3.1. Descriptive statistics
        • 3.2. Data visualization
      • 4. Data preparation and analysis
        • 4.1. Univariate feature selection
      • 5. Evaluate models
        • 5.1. Train-test split and evaluation metrics
        • 5.2. Compare models and algorithms
      • 6. Model tuning and finalizing the model
      • 7. Additional analysis: removing the volatility data
      • Conclusion
  • Case Study 3: Investor Risk Tolerance and Robo-Advisors
    • Blueprint for Modeling Investor Risk Tolerance and Enabling a Machine LearningBased Robo-Advisor
      • 1. Problem definition
      • 2. Getting startedloading the data and Python packages
        • 2.1. Loading the Python packages
        • 2.2. Loading the data
      • 3. Data preparation and feature selection
        • 3.1. Preparing the predicted variable
        • 3.2. Feature selectionlimit the feature space
        • 3.2.1. Feature elimination
      • 4. Evaluate models
        • 4.1. Train-test split
        • 4.2. Test options and evaluation metrics
        • 4.3. Compare models and algorithms
      • 5. Model tuning and grid search
      • 6. Finalize the model
        • 6.1. Results on the test dataset
      • 6.2. Feature importance and features intuition
        • 6.3. Save model for later use
      • 7. Additional step: robo-advisor dashboard
        • Input for investor characteristics
        • 7.2 Asset allocation and portfolio performance
      • Conclusion
  • Case Study 4: Yield Curve Prediction
    • Blueprint for Using Supervised Learning Models to Predict the Yield Curve
      • 1. Problem definition
      • 2. Getting startedloading the data and Python packages
        • 2.1. Loading the Python packages
        • 2.2. Loading the data
        • 3. Exploratory data analysis
        • 3.1. Descriptive statistics
        • 3.2. Data visualization
      • 4. Data preparation and analysis
      • 5. Evaluate models
        • 5.1. Train-test split and evaluation metrics
        • 5.2. Compare models and algorithms
      • 6. Model tuning and grid search.
        • Prediction comparison
      • Conclusion
  • Chapter Summary
  • Exercises
• 6. Supervised Learning: Classification
  • Case Study 1: Fraud Detection
    • Blueprint for Using Classification Models to Determine Whether a Transaction Is Fraudulent
      • 1. Problem definition
      • 2. Getting startedloading the data and Python packages
        • 2.1. Loading the Python packages
      • 3. Exploratory data analysis
        • 3.1. Descriptive statistics
        • 3.2. Data visualization
      • 4. Data preparation
      • 5. Evaluate models
        • 5.1. Train-test split and evaluation metrics
        • 5.2. Checking models
      • 6. Model tuning
        • 6.1. Model tuning by choosing the correct evaluation metric
        • 6.2. Model tuningbalancing the sample by random under-sampling
      • Conclusion
  • Case Study 2: Loan Default Probability
    • Blueprint for Creating a Machine Learning Model for Predicting Loan Default Probability
      • 1. Problem definition
      • 2. Getting startedloading the data and Python packages
        • 2.1. Loading the Python packages
        • 2.2. Loading the data
      • 3. Data preparation and feature selection
        • 3.1. Preparing the predicted variable
        • 3.2. Feature selectionlimit the feature space
        • 3.2.1. Feature elimination based on significant missing values
        • 3.2.2. Feature elimination based on intuitiveness
        • 3.2.3. Feature elimination based on the correlation
      • 4. Feature selection and exploratory analysis
        • 4.1. Feature analysis and exploration
        • 4.1.1. Analyzing the categorical features
        • 4.1.2. Analyzing the continuous features
        • 4.2. Encoding categorical data
        • 4.3. Sampling data
      • 5. Evaluate algorithms and models
        • 5.1. Train-test split
        • 5.2. Test options and evaluation metrics
        • 5.3. Compare models and algorithms
      • 6. Model tuning and grid search
      • 7. Finalize the model
        • 7.1. Results on the test dataset
        • 7.2. Variable intuition/feature importance
      • Conclusion
  • Case Study 3: Bitcoin Trading Strategy
    • Blueprint for Using Classification-Based Models to Predict Whether to Buy or Sell in the Bitcoin Market
      • 1. Problem definition
      • 2. Getting startedloading the data and Python packages
        • 2.1. Loading the Python packages
        • 2.2. Loading the data
      • 3. Exploratory data analysis
        • 3.1. Descriptive statistics
      • 4. Data preparation
        • 4.1. Data cleaning
        • 4.2. Preparing the data for classification
        • 4.3. Feature engineering
        • 4.4. Data visualization
      • 5. Evaluate algorithms and models
        • 5.1. Train-test split
        • 5.2. Test options and evaluation metrics
        • 5.3. Compare models and algorithms
        • 5.3.1. Models
      • 6. Model tuning and grid search
      • 7. Finalize the model
        • 7.1. Results on the test dataset
        • 7.2. Variable intuition/feature importance
        • 7.3. Backtesting results
      • Conclusion
  • Chapter Summary
  • Exercises
• III. Unsupervised Learning
• 7. Unsupervised Learning: Dimensionality Reduction
  • Dimensionality Reduction Techniques
    • Principal Component Analysis
      • Eigen decomposition
      • Singular value decomposition
    • Kernel Principal Component Analysis
    • t-distributed Stochastic Neighbor Embedding
  • Case Study 1: Portfolio Management: Finding an Eigen Portfolio
    • Blueprint for Using Dimensionality Reduction for Asset Allocation
      • 1. Problem definition
      • 2. Getting startedloading the data and Python packages
        • 2.1. Loading the Python packages
        • 2.2. Loading the data
      • 3. Exploratory data analysis
        • 3.1. Descriptive statistics
        • 3.2. Data visualization
      • 4. Data preparation
        • 4.1. Data cleaning
        • 4.2. Data transformation
      • 5. Evaluate algorithms and models
        • 5.1. Train-test split
        • 5.2. Model evaluation: applying principal component analysis
        • 5.2.1. Explained variance using PCA
        • 5.2.2. Looking at portfolio weights
        • 5.2.3. Finding the best eigen portfolio
        • 5.2.4. Backtesting the eigen portfolios
      • Conclusion
  • Case Study 2: Yield Curve Construction and Interest Rate Modeling
    • Blueprint for Using Dimensionality Reduction to Generate a Yield Curve
      • 1. Problem definition
      • 2. Getting startedloading the data and Python packages
        • 2.1. Loading the Python packages
        • 2.2. Loading the data
      • 3. Exploratory data analysis
        • 3.1. Descriptive statistics
        • 3.2. Data visualization
      • 4. Data preparation
        • 4.1. Data cleaning
        • 4.2. Data transformation
      • 5. Evaluate algorithms and models
        • 5.2. Model evaluationapplying principal component analysis
        • 5.2.1. Explained variance using PCA
        • 5.2.2. Intuition behind the principal components
        • 5.2.3. Reconstructing the curve using principal components
      • Conclusion
  • Case Study 3: Bitcoin Trading: Enhancing Speed and Accuracy
    • Blueprint for Using Dimensionality Reduction to Enhance a Trading Strategy
      • 1. Problem definition
      • 2. Getting startedloading the data and Python packages
        • 2.1. Loading the Python packages
      • 3. Exploratory data analysis
      • 4. Data preparation
        • 4.1. Data cleaning
        • 4.2. Preparing the data for classification
        • 4.3. Feature engineering
        • 4.4. Data visualization
      • 5. Evaluate algorithms and models
        • 5.1. Train-test split
        • 5.2. Singular value decomposition (feature reduction)
        • 5.2.1. Basic visualization of reduced features
        • 5.3. t-SNE visualization
        • 5.4. Compare models with and without dimensionality reduction
        • 5.4.1. Models
      • Conclusion
  • Chapter Summary
  • Exercises
• 8. Unsupervised Learning: Clustering
  • Clustering Techniques
    • k-means Clustering
      • k-means hyperparameters
      • Implementation in Python
    • Hierarchical Clustering
      • Implementation in Python
    • Affinity Propagation Clustering
      • Implementation in Python
  • Case Study 1: Clustering for Pairs Trading
    • Blueprint for Using Clustering to Select Pairs
      • 1. Problem definition
      • 2. Getting startedloading the data and Python packages
        • 2.1. Loading the Python packages
        • 2.2. Loading the data
      • 3. Exploratory data analysis
        • 3.1. Descriptive statistics
        • 3.2. Data visualization
      • 4. Data preparation
        • 4.1. Data cleaning
        • 4.2. Data transformation
      • 5. Evaluate algorithms and models
        • 5.1. k-means clustering
        • 5.1.1. Finding the optimal number of clusters
        • 5.1.2. Clustering and visualization
        • 5.2. Hierarchical clustering (agglomerative clustering)
        • 5.2.1. Building hierarchy graph/dendrogram
        • 5.2.2. Clustering and visualization
        • 5.3. Affinity propagation
        • 5.4. Cluster evaluation
        • Visualizing the return within a cluster
      • 6. Pairs selection
        • Cointegration and pair selection function
      • Conclusion
  • Case Study 2: Portfolio Management: Clustering Investors
    • Blueprint for Using Clustering for Grouping Investors
      • 1. Problem definition
      • 2. Getting startedloading the data and Python packages
        • 2.1. Loading the Python packages
        • 2.2. Loading the data
      • 3. Exploratory data analysis
        • 3.1. Descriptive statistics
        • 3.2. Data visualization
      • 4. Data preparation
        • 4.1. Data cleaning
        • 4.2. Data transformation
      • 5. Evaluate algorithms and models
        • 5.1. k-means clustering
        • 5.1.1. Finding the optimal number of clusters
        • 5.1.2. Clustering and visualization
        • 5.2. Affinity propagation
        • 5.3. Cluster evaluation
      • 6. Cluster intuition
      • Conclusion
  • Case Study 3: Hierarchical Risk Parity
    • Blueprint for Using Clustering to Implement Hierarchical Risk Parity
      • 1. Problem definition
      • 2. Getting startedloading the data and Python packages
        • 2.1. Loading the Python packages
      • 3. Exploratory data analysis
      • 4. Data preparation
        • 4.1. Data cleaning
        • 4.2. Data transformation
      • 5. Evaluate algorithms and models
        • 5.1. Building a hierarchy graph/dendrogram
        • 5.2. Steps for hierarchical risk parity
        • 5.2.1. Quasi-diagonalization
        • 5.2.2. Recursive bisection
        • 5.3. Comparison against other asset allocation methods
        • 5.4. Getting the portfolio weights for all types of asset allocation
      • 6. Backtesting
        • In-sample results
        • Out-of-sample results
      • Conclusion
  • Chapter Summary
  • Exercises
• IV. Reinforcement Learning and Natural Language Processing
• 9. Reinforcement Learning
  • Reinforcement LearningTheory and Concepts
    • RL Components
      • Policy
      • Value function (and Q-value)
      • Model
      • RL components in a trading context
    • RL Modeling Framework
      • Bellman equations
      • Markov decision processes
      • Temporal difference learning
      • Artificial neural network and deep learning
    • Reinforcement Learning Models
      • Model-based algorithms
      • Model-free algorithms
      • Q-Learning
      • SARSA
      • Deep Q-Network
      • Policy gradient
    • Key Challenges in Reinforcement Learning
  • Case Study 1: Reinforcement LearningBased Trading Strategy
    • Blueprint for Creating a Reinforcement LearningBased Trading Strategy
      • 1. Problem definition
      • 2. Getting startedloading the data and Python packages
        • 2.1. Loading the Python packages
        • 2.2. Loading the data
      • 3. Exploratory data analysis
      • 4. Data preparation
        • 4.1. Data cleaning
      • 5. Evaluate algorithms and models
        • 5.1. Train-test split
        • 5.2. Implementation steps and modules
      • Modules and functions
        • 5.3. Agent class
        • 5.4. Helper functions
        • 5.5. Training the model
        • 5.6. Model tuning
      • 6. Testing the data
      • Conclusion
  • Case Study 2: Derivatives Hedging
    • Blueprint for Implementing a Reinforcement LearningBased Hedging Strategy
      • 1. Problem definition
      • 2. Getting started
        • 2.1. Loading the Python packages
        • 2.2. Generating the data
      • 3. Exploratory data analysis
      • 4. Evaluate algorithms and models
        • 4.1. Policy gradient script
        • 4.2. Training the data
      • 5. Testing the data
        • 5.1. Helper functions for comparison against Black-Scholes
        • 5.1.1. Black-Scholes price and delta
        • 5.1.2. Test results and plotting
        • 5.1.3. Hedging error for Black-Scholes replication
        • 5.2. Comparison between Black-Scholes and reinforcement learning
        • 5.2.1. Test at 99% risk aversion
        • 5.2.2. Changing moneyness
        • 5.2.3. Changing drift
        • 5.2.4. Shifted volatility
      • Conclusion
  • Case Study 3: Portfolio Allocation
    • Blueprint for Creating a Reinforcement LearningBased Algorithm for Portfolio Allocation
      • 1. Problem definition
      • 2. Getting startedloading the data and Python packages
        • 2.1. Loading the Python packages
        • 2.2. Loading the data
      • 3. Exploratory data analysis
        • 3.1. Descriptive statistics
      • 4. Evaluate algorithms and models
        • 4.1. Agent and cryptocurrency environment script
        • 4.3. Training the data
      • 5. Testing the data
      • Conclusion
  • Chapter Summary
  • Exercises
• 10. Natural Language Processing
  • Natural Language Processing: Python Packages
    • NLTK
    • TextBlob
    • spaCy
  • Natural Language Processing: Theory and Concepts
    • 1. Preprocessing
      • 1.1. Tokenization
      • 1.2. Stop words removal
      • 1.3. Stemming
      • 1.4. Lemmatization
      • 1.5. PoS tagging
      • 1.6. Named entity recognition
        • 1.7. spaCy: All of the above steps in one go
    • 2. Feature Representation
      • 2.1. Bag of wordsword count
      • 2.2. TF-IDF
      • 2.3. Word embedding
      • 2.3.1. Pretrained model: Via spaCy
      • 2.3.2. Pretrained model: Word2Vec using gensim package
    • 3. Inference
      • 3.1. Supervised learning exampleNaive Bayes
      • 3.2. Unsupervised learning example: LDA
  • Case Study 1: NLP and Sentiment AnalysisBased Trading Strategies
    • Blueprint for Building a Trading Strategy Based on Sentiment Analysis
      • 1. Problem definition
      • 2. Getting startedloading the data and Python packages
        • 2.1. Loading the Python packages
        • 2.2. Loading the data
      • 3. Data preparation
        • 3.1. Preprocessing news data
      • 4. Evaluate models for sentiment analysis
        • 4.1. Predefined modelTextBlob package
      • 4.2. Supervised learningclassification algorithms and LSTM
      • 4.3. Unsupervisedmodel based on a financial lexicon
      • 4.4. Exploratory data analysis and comparison
      • 5. Models evaluationbuilding a trading strategy
        • 5.1. Setting up a strategy
        • 5.2. Results for individual stocks
        • 5.3. Results for multiple stocks
      • 5.4. Varying the strategy time period
      • Conclusion
  • Case Study 2: Chatbot Digital Assistant
    • Blueprint for Creating a Custom Chatbot Using NLP
      • 1. Problem definition
      • 2. Getting startedloading the libraries
        • 2.1. Load libraries
      • 3. Training a default chatbot
      • 4. Data preparation: Customized chatbot
        • 4.1. Data construction
        • 4.2. Training data
        • 5. Model creation and training
      • 5.1. Model optimization function
      • 5.2. Custom logic adapter
      • 5.3. Model usagetraining and testing
      • Conclusion
  • Case Study 3: Document Summarization
    • Blueprint for Using NLP for Document Summarization
      • 1. Problem definition
      • 2. Getting startedloading the data and Python packages
        • 2.1. Loading the Python packages
      • 3. Data preparation
      • 4. Model construction and training
      • 5. Visualization of topics
        • 5.1. Topic visualization
        • 5.2. Word cloud
      • Conclusion
  • Chapter Summary
  • Exercises
• Index

pokaż cały spis treści

ukryj recenzje