ANN Trend PredictionThis trend indicator utilizes an artificial neural network (ANN) to predict the next market reversal within a certain range of previous candles. The larger the range of previous candles you set, the fewer reversals will be predicted, and trends will tend to last longer.
The ANN is trained on the BTCUSD 4-hour chart, so using it on other assets or timeframes may yield suboptimal results. It takes three input values: the closing price, the Stochastic RSI, and a Choppiness Indicator. Based on these inputs, the ANN categorizes the current candle as part of an uptrend, downtrend, or as undefined.
Compared to an EMA-based trend indicator, this ANN identifies reversals several candles earlier. It achieves this by detecting subtle patterns in the input values that typically appear before a market turnaround. These patterns are somewhat specific to that chosen asset and timeframe.
The results are displayed using rows of triangles that indicate the predicted price direction. The price levels of the triangles correspond to the closing price at the last reversal. The area between the triangle row and the price is colored green if the ANN correctly predicted the move, and red if it did not.
This indicator is designed to showcase the capabilities and potential of ANNs, and is not intended for actual trading use. The ANN can be trained on any other input values, assets and timeframes for several predictions tasks.
You can use the Predicted_Trend_Signal of this Indicator in any backtest indicator. In the Backtester just grap the Predicted_Trend_Signal. downtrend = 1, uptrend = -1, undefined = 0
Feel free to write me a comment.
Ml
Machine Learning Cross-Validation Split & Batch HighlighterThis indicator is designed for traders and analysts who employ Machine Learning (ML) techniques for cross-validation in financial markets.
The script visually segments a selected range of historical price data into splits and batches, helping in the assessment of model performance over different market conditions.
User
Theory
In ML, cross-validation is a technique to assess the generalizability of a model, typically by partitioning the data into a set of "folds" or "splits." Each split acts as a validation set, while the others form the training set. This script takes a unique approach by considering the sequential nature of financial time series data, where random shuffling of data (as in traditional cross-validation) can disrupt the temporal order, leading to misleading results.
Chronological Integrity of Splits
Even if the order of the splits is shuffled for cross-validation purposes, the data within each split remains in its original chronological sequence. This feature is crucial for time series analysis, as it respects the inherent order-dependency of financial markets. Thus, each split can be considered a microcosm of market behavior, maintaining the integrity of trends, cycles, and patterns that could be disrupted by random sampling.
The script allows users to define the number of splits and the size of each batch within a split. By doing so, it maintains the chronological sequence of the data, ensuring that the validation set is representative of a future time period that the model would predict.
www.tradingview.com
Parameters
Number of Splits: Defines how many segments the selected data range will be divided into. Each split serves as a standalone testing ground for the ML model. (Up to 24)
Batch Size: Determines the number of bars (candles) in each batch within a split. Smaller batches can help pinpoint overfitting at a finer granularity.
Start Index: The bar index from where the historical data range begins. It sets the starting point for data analysis.
End Index: The bar index where the historical data range ends. It marks the cutoff for data to be included in the model assessment.
Usage
To use this script effectively:
1 - Input the Start Index and End Index to define the historical data range you wish to analyze.
2 - Adjust the Number of Splits to create multiple validation sets for cross-validation.
3 - Set the Batch Size to control the granularity of each validation set within the splits.
4 - The script will highlight the background of each batch within the splits using alternating shades, allowing for a clear visual distinction of the data segmentation.
By maintaining the temporal sequence and allowing for adjustable granularity, the "ML Split and Batch Highlighter" aids in creating a robust validation framework for time series forecasting models in finance.
ML - Momentum Index (Pivots)Building upon the innovative foundations laid by Zeiierman's Machine Learning Momentum Index (MLMI), this variation introduces a series of refinements and new features aimed at bolstering the model's predictive accuracy and responsiveness. Licensed under the Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License (CC BY-NC-SA 4.0), my adaptation seeks to enhance the original by offering a more nuanced approach to momentum-based trading.
Key Features :
Pivot-Based Analysis: Shifting focus from trend crosses to pivot points, this version employs pivot bars to offer a distinct perspective on market momentum, aiding in the identification of critical reversal points.
Extended Parameter Set: By integrating additional parameters for making predictions, the model gains improved adaptability, allowing for finer tuning to match market conditions.
Dataset Size Limitation: To ensure efficiency and mitigate the risk of calculation timeouts, a cap on the dataset size has been implemented, balancing between comprehensive historical analysis and computational agility.
Enhanced Price Source Flexibility: Users can select between closing prices or (suggested) OHLC4 as the basis for calculations, tailoring the indicator to different analysis preferences and strategies.
This adaptation not only inherits the robust framework of the original MLMI but also introduces innovations to enhance its utility in diverse trading scenarios. Whether you're looking to refine your short-term trading tactics or seeking stable indicators for long-term strategies, the ML - Momentum Index (Pivots) offers a versatile tool to navigate the complexities of the market.
For a deeper understanding of the modifications and to leverage the full potential of this indicator, users are encouraged to explore the tooltips and documentation provided within the script.
The Momentum Indicator calculations have been transitioned to the MLMomentumIndex library, simplifying the process of integration. Users can now seamlessly incorporate the momentumIndexPivots function into their scripts to conduct detailed momentum analysis with ease.
Machine Learning using Neural Networks | EducationalThe script provided is a comprehensive illustration of how to implement and execute a simplistic Neural Network (NN) on TradingView using PineScript.
It encompasses the entire workflow from data input, weight initialization, implicit neuron calculation, feedforward computation, backpropagation for weight adjustments, generating predictions, to visualizing the Mean Squared Error (MSE) Loss Curve for monitoring the training phase.
In the visual example above, you can see that the prediction is not aligned with the actual value. This is intentional for demonstrative purposes, and by incrementing the Epochs or Learning Rate, you will see these two values converge as the accuracy increases.
Hyperparameters:
Learning Rate, Epochs, and the choice between Simple Backpropagation and a verbose version are declared as script inputs, allowing users to tailor the training process.
Initialization:
Random initialization of weight matrices (w1, w2) is performed to ensure asymmetry, promoting effective gradient updates. A seed is added for reproducibility.
Utility Functions:
Functions for matrix randomization, sigmoid activation, MSE loss calculation, data normalization, and standardization are defined to streamline the computation process.
Neural Network Computation:
The feedforward function computes the hidden and output layer values given the input.
Two variants of the backpropagation function are provided for weight adjustment, with one offering a more verbose step-by-step computation of gradients.
A wrapper train_nn function iterates through epochs, performing feedforward, loss computation, and backpropagation in each epoch while logging and collecting loss values.
Training Invocation:
The input data is prepared by normalizing it to a value between 0 and 1 using the maximum standardized value, and the training process is invoked only on the last confirmed bar to preserve computational resources.
Output Forecasting and Visualization:
Post training, the NN's output (predicted price) is computed, standardized and visualized alongside the actual price on the chart.
The MSE loss between the predicted and actual prices is visualized, providing insight into the prediction accuracy.
Optionally, the MSE Loss Curve is plotted on the chart, illustrating the loss trajectory through epochs, assisting in understanding the training performance.
Customizable Visualization:
Various inputs control visualization aspects like Chart Scaling, Chart Horizontal Offset, and Chart Vertical Offset, allowing users to adapt the visualization to their preference.
-------------------------------------------------------
The following is this Neural Network structure, consisting of one hidden layer, with two hidden neurons.
Through understanding the steps outlined in my code, one should be able to scale the NN in any way they like, such as changing the input / output data and layers to fit their strategy ideas.
Additionally, one could forgo the backpropagation function, and load their own trained weights into the w1 and w2 matrices, to have this code run purely for inference.
-------------------------------------------------------
While this demonstration does create a “prediction”, it is on historical data. The purpose here is educational, rather than providing a ready tool for non-programmer consumers.
Normally in Machine Learning projects, the training process would be split into two segments, the Training and the Validation parts. For the purpose of conveying the core concept in a concise and non-repetitive way, I have foregone the Validation part. However, it is merely the application of your trained network on new data (feedforward), and monitoring the loss curve.
Essentially, checking the accuracy on “unseen” data, while training it on “seen” data.
-------------------------------------------------------
I hope that this code will help developers create interesting machine learning applications within the Tradingview ecosystem.
Relational Quadratic Kernel Channel [Vin]The Relational Quadratic Kernel Channel (RQK-Channel-V) is designed to provide more valuable potential price extremes or continuation points in the price trend.
Example:
Usage:
Lookback Window: Adjust the "Lookback Window" parameter to control the number of previous bars considered when calculating the Rational Quadratic Estimate. Longer windows capture longer-term trends, while shorter windows respond more quickly to price changes.
Relative Weight: The "Relative Weight" parameter allows you to control the importance of each data point in the calculation. Higher values emphasize recent data, while lower values give more weight to historical data.
Source: Choose the data source (e.g., close price) that you want to use for the kernel estimate.
ATR Length: Set the length of the Average True Range (ATR) used for channel width calculation. A longer ATR length results in wider channels, while a shorter length leads to narrower channels.
Channel Multipliers: Adjust the "Channel Multiplier" parameters to control the width of the channels. Higher multipliers result in wider channels, while lower multipliers produce narrower channels. The indicator provides three sets of channels, each with its own multiplier for flexibility.
Details:
Rational Quadratic Kernel Function:
The Rational Quadratic Kernel Function is a type of smoothing function used to estimate a continuous curve or line from discrete data points. It is often used in time series analysis to reduce noise and emphasize trends or patterns in the data.
The formula for the Rational Quadratic Kernel Function is generally defined as:
K(x) = (1 + (x^2) / (2 * α * β))^(-α)
Where:
x represents the distance or difference between data points.
α and β are parameters that control the shape of the kernel. These parameters can be adjusted to control the smoothness or flexibility of the kernel function.
In the context of this indicator, the Rational Quadratic Kernel Function is applied to a specified source (e.g., close prices) over a defined lookback window. It calculates a smoothed estimate of the source data, which is then used to determine the central value of the channels. The kernel function allows the indicator to adapt to different market conditions and reduce noise in the data.
The specific parameters (length and relativeWeight) in your indicator allows to fine-tune how the Rational Quadratic Kernel Function is applied, providing flexibility in capturing both short-term and long-term trends in the data.
To know more about unsupervised ML implementations, I highly recommend to follow the users, @jdehorty and @LuxAlgo
Optimizing the parameters:
Lookback Window (length): The lookback window determines how many previous bars are considered when calculating the kernel estimate.
For shorter-term trading strategies, you may want to use a shorter lookback window (e.g., 5-10).
For longer-term trading or investing, consider a longer lookback window (e.g., 20-50).
Relative Weight (relativeWeight): This parameter controls the importance of each data point in the calculation.
A higher relative weight (e.g., 2 or 3) emphasizes recent data, which can be suitable for trend-following strategies.
A lower relative weight (e.g., 1) gives more equal importance to historical and recent data, which may be useful for strategies that aim to capture both short-term and long-term trends.
ATR Length (atrLength): The length of the Average True Range (ATR) affects the width of the channels.
Longer ATR lengths result in wider channels, which may be suitable for capturing broader price movements.
Shorter ATR lengths result in narrower channels, which can be helpful for identifying smaller price swings.
Channel Multipliers (channelMultiplier1, channelMultiplier2, channelMultiplier3): These parameters determine the width of the channels relative to the ATR.
Adjust these multipliers based on your risk tolerance and desired channel width.
Higher multipliers result in wider channels, which may lead to fewer signals but potentially larger price movements.
Lower multipliers create narrower channels, which can result in more frequent signals but potentially smaller price movements.
Machine Learning Regression Trend [LuxAlgo]The Machine Learning Regression Trend tool uses random sample consensus (RANSAC) to fit and extrapolate a linear model by discarding potential outliers, resulting in a more robust fit.
🔶 USAGE
The proposed tool can be used like a regular linear regression, providing support/resistance as well as forecasting an estimated underlying trend.
Using RANSAC allows filtering out outliers from the input data of our final fit, by outliers we are referring to values deviating from the underlying trend whose influence on a fitted model is undesired. For financial prices and under the assumptions of segmented linear trends, these outliers can be caused by volatile moves and/or periodic variations within an underlying trend.
Adjusting the "Allowed Error" numerical setting will determine how sensitive the model is to outliers, with higher values returning a more sensitive model. The blue margin displayed shows the allowed error area.
The number of outliers in the calculation window (represented by red dots) can also be indicative of the amount of noise added to an underlying linear trend in the price, with more outliers suggesting more noise.
Compared to a regular linear regression which does not discriminate against any point in the calculation window, we see that the model using RANSAC is more conservative, giving more importance to detecting a higher number of inliners.
🔶 DETAILS
RANSAC is a general approach to fitting more robust models in the presence of outliers in a dataset and as such does not limit itself to a linear regression model.
This iterative approach can be summarized as follow for the case of our script:
Step 1: Obtain a subset of our dataset by randomly selecting 2 unique samples
Step 2: Fit a linear regression to our subset
Step 3: Get the error between the value within our dataset and the fitted model at time t , if the absolute error is lower than our tolerance threshold then that value is an inlier
Step 4: If the amount of detected inliers is greater than a user-set amount save the model
Repeat steps 1 to 4 until the set number of iterations is reached and use the model that maximizes the number of inliers
🔶 SETTINGS
Length: Calculation window of the linear regression.
Width: Linear regression channel width.
Source: Input data for the linear regression calculation.
🔹 RANSAC
Minimum Inliers: Minimum number of inliers required to return an appropriate model.
Allowed Error: Determine the tolerance threshold used to detect potential inliers. "Auto" will automatically determine the tolerance threshold and will allow the user to multiply it through the numerical input setting at the side. "Fixed" will use the user-set value as the tolerance threshold.
Maximum Iterations Steps: Maximum number of allowed iterations.
Universal Moving Average Convergence DivergenceI changed MACD formula to divergence of (MA26/MA12 - 1).
And its make it more useful.
Cuz:
1) comparability with all other coins with different prices.
2) fix small numbers in low price coines like shiba
3) making a good indicator like RSI to use it for optimization and ML/AI projects as a variable
Most important thing about this indicator is that its Universal
Now you can compare the UMACD of Shiba with Bitcoin without any problem in matamatics space.No need to use virtuality and its important in Optimization problems that we rediuse the problem from a picture to a number(A plot to a list of numbers)
If we don't care about exagrated pumps and dumps, we can say to it Normalized-MACD too. Cuz in normal situations its MAX ≈ 0.1 and MIN ≈ -0.1
WIPNNetworkLibrary "WIPNNetwork"
this is a work in progress (WIP) and prone to have some errors, so use at your own risk...
let me know if you find any issues..
Method for a generalized Neural Network.
network(x) Generalized Neural Network Method.
Parameters:
x : TODO: add parameter x description here
Returns: TODO: add what function returns
FunctionNNLayerLibrary "FunctionNNLayer"
Generalized Neural Network Layer method.
function(inputs, weights, n_nodes, activation_function, bias, alpha, scale) Generalized Layer.
Parameters:
inputs : float array, input values.
weights : float array, weight values.
n_nodes : int, number of nodes in layer.
activation_function : string, default='sigmoid', name of the activation function used.
bias : float, default=1.0, bias to pass into activation function.
alpha : float, default=na, if required to pass into activation function.
scale : float, default=na, if required to pass into activation function.
Returns: float
FunctionNNPerceptronLibrary "FunctionNNPerceptron"
Perceptron Function for Neural networks.
function(inputs, weights, bias, activation_function, alpha, scale) generalized perceptron node for Neural Networks.
Parameters:
inputs : float array, the inputs of the perceptron.
weights : float array, the weights for inputs.
bias : float, default=1.0, the default bias of the perceptron.
activation_function : string, default='sigmoid', activation function applied to the output.
alpha : float, default=na, if required for activation.
scale : float, default=na, if required for activation.
@outputs float
MLActivationFunctionsLibrary "MLActivationFunctions"
Activation functions for Neural networks.
binary_step(value) Basic threshold output classifier to activate/deactivate neuron.
Parameters:
value : float, value to process.
Returns: float
linear(value) Input is the same as output.
Parameters:
value : float, value to process.
Returns: float
sigmoid(value) Sigmoid or logistic function.
Parameters:
value : float, value to process.
Returns: float
sigmoid_derivative(value) Derivative of sigmoid function.
Parameters:
value : float, value to process.
Returns: float
tanh(value) Hyperbolic tangent function.
Parameters:
value : float, value to process.
Returns: float
tanh_derivative(value) Hyperbolic tangent function derivative.
Parameters:
value : float, value to process.
Returns: float
relu(value) Rectified linear unit (RELU) function.
Parameters:
value : float, value to process.
Returns: float
relu_derivative(value) RELU function derivative.
Parameters:
value : float, value to process.
Returns: float
leaky_relu(value) Leaky RELU function.
Parameters:
value : float, value to process.
Returns: float
leaky_relu_derivative(value) Leaky RELU function derivative.
Parameters:
value : float, value to process.
Returns: float
relu6(value) RELU-6 function.
Parameters:
value : float, value to process.
Returns: float
softmax(value) Softmax function.
Parameters:
value : float array, values to process.
Returns: float
softplus(value) Softplus function.
Parameters:
value : float, value to process.
Returns: float
softsign(value) Softsign function.
Parameters:
value : float, value to process.
Returns: float
elu(value, alpha) Exponential Linear Unit (ELU) function.
Parameters:
value : float, value to process.
alpha : float, default=1.0, predefined constant, controls the value to which an ELU saturates for negative net inputs. .
Returns: float
selu(value, alpha, scale) Scaled Exponential Linear Unit (SELU) function.
Parameters:
value : float, value to process.
alpha : float, default=1.67326324, predefined constant, controls the value to which an SELU saturates for negative net inputs. .
scale : float, default=1.05070098, predefined constant.
Returns: float
exponential(value) Pointer to math.exp() function.
Parameters:
value : float, value to process.
Returns: float
function(name, value, alpha, scale) Activation function.
Parameters:
name : string, name of activation function.
value : float, value to process.
alpha : float, default=na, if required.
scale : float, default=na, if required.
Returns: float
derivative(name, value, alpha, scale) Derivative Activation function.
Parameters:
name : string, name of activation function.
value : float, value to process.
alpha : float, default=na, if required.
scale : float, default=na, if required.
Returns: float
MLLossFunctionsLibrary "MLLossFunctions"
Methods for Loss functions.
mse(expects, predicts) Mean Squared Error (MSE) " MSE = 1/N * sum ((y - y')^2) ".
Parameters:
expects : float array, expected values.
predicts : float array, prediction values.
Returns: float
binary_cross_entropy(expects, predicts) Binary Cross-Entropy Loss (log).
Parameters:
expects : float array, expected values.
predicts : float array, prediction values.
Returns: float
neutronix community bot ML + Alerts 4h-daily (mod. capissimo)Gm traders,
i have been a python programmer for some years studying artificial intelligence for general purpose; after some time i finally decided to have a look at some finance related stuff and scripts.
Moved by curiosity i've decided to make some but decisive modifications to a script i tried to use initially but without success: the LVQ machine learning strategy.
So after studying the charts and indicators, i have rewritten this script made by Capissimo and added heavy filtering thanks to vwap and vwma, then fixed repaint and other issues.
I hope you enjoy it and that it could increase your possibilities of success in trading.
HOW TO USE THE SCRIPT
Add the script to 3h+ charts like for example BTC 4h, 6h, 8h, 12h, daily. (In order for it to work on shorter timeframes charts you can try to change to lookback window but i dont advise it).
Change only rsi and volfilter(volume filtering) settings to try to find the best winrate. Leave dataset to open. Fyi the winrate isn't 100% accurate but can give you a raw vision of final results.
Use alerts included for trading and and in options click on 'Once per bar'. If you have checked 'Reverse Signals' in the control panel you have got more 'risky' signals so be advised if trading futures and stocks.
Exit trade signals not provided, so it is recommended the use of take profits and stop loss (1.5:1 ratio)
As always, the script is for study purposes. Do not risk more than you can spend!
Original LVQ-based strategy made by capissimo
Modified by gravisxv 13/10/2021
Minkowski Distance Factor Adaptive Period MACDHi, this script comes from the idea that Ricardo Santos' Minkovski Distance Function is transferred to the period as a factor.
Minkowski distance is used as a percentage factor with the help of Relative Strength Index function.
Minkowski Distance Function Script :
And thus an adaptive MACD was created.
This script can give much better results in more optimized larger periods.
I leave the decision to determine the periods and weights.
I used the weights of 9,12,26 and periods created with multiplied by factor.
Regards.
CBOE PCR Factor Dependent Variable Odd Generator This script is the my Dependent Variable Odd Generator script :
with the Put / Call Ratio ( PCR ) appended, only for CBOE and the instruments connected to it.
For CBOE this script is more accurate and faster than Dependent Variable Odd Generator. And the stagnant market odds are better and more realistic.
Do not use for timeframe periods less than 1 day.
Because PCR data may give repaint error.
My advice is to use the 1-week bars to gain insight into your analysis.
This code is open source under the MIT license. If you have any improvements or corrections to suggest, please send me a pull request via the github repository github.com
I hope it will help your work.Best regards!
