/Deep-Learning-Machine-Learning-Stock

Deep Learning and Machine Learning stocks represent a promising long-term or short-term opportunity for investors and traders.

Primary LanguageJupyter NotebookMIT LicenseMIT

Deep Learning and Machine Learning for Stock Predictions

Description: This is for learning, studying, researching, and analyzing stock in deep learning (DL) and machine learning (ML). Predicting Stock with Machine Learning method or Deep Learning method with different types of algorithm. Experimenting in stock data to see how it works and why it works or why it does not work that way. Using different types of stock strategies in machine learning or deep learning. Using Technical Analysis or Fundamental Analysis in machine learning or deep learning to predict the future stock price. In addition, to predict stock in long terms or short terms.

Machine learning is a subset of artificial intelligence involved with the creating of algorithms that can change itself without human intervention to produce an output by feeding itself through structured data. On the other hand, deep learning is a subset of machine learning where algorithms created, but the function are like machine learning and many of the different type of algorithms give a different interpretation of the data. The network of algorithms called artificial neural networks and is similar to neural connections that exist in the human brain.

Three main types of data: Categorical, Discrete, and Continuous variables

  1. Categorical variable(Qualitative): Label data or distinct groups.
    Example: location, gender, material type, payment, highest level of education
  2. Discrete variable (Class Data): Numerica variables but the data is countable number of values between any two values.
    Example: customer complaints or number of flaws or defects, Children per Household, age (number of years)
  3. Continuous variable (Quantitative): Numeric variables that have an infinite number of values between any two values. Example: length of a part or the date and time a payment is received, running distance, age (infinitly accurate and use an infinite number of decimal places)

Data Use

  1. For 'Quantitative data' is used with all three centre measures (mean, median and mode) and all spread measures.
  2. For 'Class data' is used with median and mode.
  3. For 'Qualitative data' is for only with mode.

Two types of problems:

  1. Classification (predict label)
  2. Regression (predict values)

Bias-Variance Tradeoff

Bias

  • Bias is the difference between our actual and predicted values.
  • Bias is the simple assumptions that our model makes about our data to be able to predict new data.
  • Assumptions made by a model to make a function easier to learn.

Variance

  • Variance is opposite of bias.
  • Variance is variability of model prediction for a given data point or a value that tells us the spread of our data.
  • If you train your data on training data and obtain a very low error, upon changing the data and then training the same.

Overfitting, Underfitting, and the bias-variance tradeoff

Overfitted is when the model memorizes the noise and fits too closely to the training set. Good fit is a model that learns the training dataset and genernalizes well with the old out dataset. Underfitting is when it cannot establish the dominant trend within the data; as a result, in training errors and poor performance of the model.

Overfitting:

Overfitting model is a good model with the training data that fit or at lease with near each observation; however, the model mist the point and random noise is capture inside the model. The model have low training error and high CV error, low in-sample error and high out-of-sample error, and high variance.

  1. High Train Accuracy
  2. Low Test Accuracy

Avoiding Overfitting:

  1. Early stopping - stop the training before the model starts learning the noise within the model.
  2. Training with more data - adding more data will increase the accuracy of the modelor can help algorithms detect the signal better.
  3. Data augmentation - add clean and relevant data into training data.
  4. Feature selection - Use important features within the data. Remove features.
  5. Regularization - reduce features by using regularization methods such as L1 regularization, Lasso regularization, and dropout.
  6. Ensemble methods - combine predictions from multiple separate models such as bagging and boosting.
  7. Increase training data.

Good fit:

Good fit:

  1. High Train Accuracy
  2. High Test Accuracy

Underfitting:

Underfitting model is not perfect, so it does not capture the underlying logic of the data. Therefore, the model does not have strong predictive power with low accuracy. The model have large training set error, large in-sample error, and high bias.

  1. Low Train Accuracy
  2. Low Test Accuracy

Avoiding Underfitting:

  1. Decrease regularization - reduce the variance with a model by applying a penalty to the input parameters with the larger coefficients such as L1 regularization, Lasso regularization, dropout, etc.
  2. Increase the duration of training - extending the duration of training because stopping the training early will cause underfit model.
  3. Feature selection - not enough predictive features present, then adding more features or features with greater importance would improve the model.
  4. Increase the number of features - performing feature engineering
  5. Remove noise from the data

Python Reviews

Step 1 through step 8 is a reviews in python.
After step 8, everything you need to know that is relate to data analysis, data engineering, data science, machine learning, and deep learning.

List of Machine Learning Algorithms for Stock Trading

Most Common Regression Algorithms

  1. Simple Linear Regression Model
  2. Logistic Regression
  3. Lasso Regression
  4. Support Vector Machines
  5. Polynomial Regression
  6. Stepwise Regression
  7. Ridge Regression
  8. Multivariate Regression Algorithm
  9. Multiple Regression Algorithm
  10. K Means Clustering Algorithm
  11. Naïve Bayes Classifier Algorithm
  12. Random Forests
  13. Decision Trees
  14. Nearest Neighbours
  15. Lasso Regression
  16. ElasticNet Regression
  17. Reinforcement Learning
  18. Artificial Intelligence
  19. MultiModal Network
  20. Biologic Intelligence

Different Types of Machine Learning Algorithms and Models

Algorithms is a process and set of instructions to solve a class of problems. In addition, algorithms perform a computation such as calculations, data processing, automated reasoning, and other tasks. A machine learning algorithms is a method that provides the systems to have the ability to automatically learn and improve from experience without being formulated.

Prerequistes

Python 3.5+
Jupyter Notebook Python 3

🔲 Add more of algorithms and different types of algorithms

Authors

* Tin Hang

Disclaimer

🔻 Do not use this code for investing or trading in the stock market. However, if you are interest in the stock market, you should read 📚 books that relate to stock market, investment, or finance. On the other hand, if you into quant or machine learning, read books about 📘 machine trading, algorithmic trading, and quantitative trading. You should read 📗 about Machine Learning and Deep Learning to understand the concept, theory, and the mathematics. On the other hand, you should read academic paper and do research online about machine learning and deep learning on 💻