$SSFT Short term price action until $1Sonasoft corp was first discovered by retail traders in January and since then it's been rising steadily. Current bear market brought it down about 40% since recent high at 0.40 but we know that this is a temporary dip. As the market gets back on its feet SSFT will continue on a similar path that I've drawn on the chart.
Sonasoft was not really aware of this increasing investor interest but recent new PR showed that traders are demanding more communication and transparency from the company and Sonasoft will be increasing its communication and marketing in the coming weeks.
"Shareholders can expect the following IR initiatives to be implemented by the end of the month:
1. Updated IR page: We will be updating the Investors Relations page on our website. The new IR section will include quarterly and annual financial reports, SEC reports, press releases, investor presentations, IR contact information, stock information such as stock quotes, charts, and other resources as they become available. We will also include a sign-up form for investors to receive periodic information via emails such as newsletters, blog posts, updates, and notifications of upcoming events.
2. More frequent corporate communications from the Company relating to material events and business developments.
3. Increased media presence and management to make the time for interviews/articles with various outlets.
4. The Company intends to begin hosting Quarterly shareholder conference calls later this year."
If you consider that it's gone up this much only by investors' research, upcoming PR as in news, deals, financial reports will have multiplying effect.
Company has ongoing client agreements with Google and Fidelity that hasn't been reflected in the financial reports yet.
They are adding new team members in the AI engineering department and have a working Proof of Concept for their flagship product NUGENE with Delaware Electric Company.
The future of Sonasoft is expansion along with the AI market and finally an acquisition by one of the giants once it establishes key technology and client list firmly.
Company is in good financial condition that's why this a safe investment among risky penny stocks.
Currently it's oversold on 4h chartwith VIX indicator showing the bottom and CMF divergence formed at this level.
First resistance is at recent high at $0.40 then comes the previous high at $0.49-0.50
after that fibo 1.618 falls at 0.78 when 0.5 is broken. And finally $1
I expect all of this to happen in the span of 2 months. If market continues to steadily rise until then, SSFT will be on of the best stocks benefiting from that influx of cash in to the market.
Please do your own DD this is just some technical analysis to predict a possible scenario.
Neuralnetwork
Neural Network CrawlYes, this indicator has been out for a while as well. Sorry if you haven't seen it because I haven't posted an idea.
This is available to everyone and can be found in your indicators on your chart.
About the Indicator
The Neural Network Crawl is a unique algorithm that identifies clusters of random walks that are crossing above or below the market price of the asset.
The random walks always exist, but the specific series that contribute to the cluster can only be seen during their significant period.
When the price trends strongly in a direction, it is more likely that it will traverse a significant amount of random walks and form a cluster.
The random walks are derived from a random selection of logarithmic movements in the last 200 bars and have been spawned at the beginning of price history.
Additionally, if you add the indicator to your chart multiple times and change the modifier in the settings panel, you can view more random walks that contribute to the clusters as seen in the screenshot below.
Cheers!
CADCHF. Price will not be subject to strong changes.Hello dear subscribers!
The price of the Canadian dollar will be kept in the range zone and will not leave its borders soon.
The price will not undergo strong changes,
so trading for a profit with this currency is possible,
but not very profitable as we would like.
If you liked this idea, please like and subscribe to my profile.
This idea does not provide the financial advice.
Development Log for Neural Network PrototypeThe idea, at the core:
Port a limited RNN/LSTM Neural Network model from Python with a reduced training set and dimension size for layers to demonstrate that a fully functional (even if limited) Neural Net can work in Pine.
Limited model + having the python code on hand = Able to test and verify components in Pine at every step, in theory
The model/script I'm attempting to implement a limited subset of is detailed here:
iamtrask.github.io
A dataset in binary is required, but binary does not exist in pinescript, thus:
To do this, decimal to binary and binary to decimal functions are required. This didn't exist previously - I've written a script to accomplish just that:
Originally, this was going to have a input_dim of 2, hidden_dim of 16, but I've changed the hidden_dim to 8 (binary dimensions from 8 to 5) to reduce the dataset range to max 32 while I figure out to implement working pseudo-arrays and state updates. I've looked at RicardoSantos's scripts for Markov and Pseudoarrays, and will be using them as a reference going forward.
I've verified the output of the Sigmoid function and 1st derivative of the Sigmoid function in Python for values of (-1,0,0.5,1 ). I've yet to publish the Sigmoid script pending approval from TV moderators about including python code that is commented out at the bottom to verify the results of that script.
What I'm trying to do here with training dataset generation was unsuccessful, for multiple reasons:
Lack of formal array constructs in pine
Psuedorandom Number generator limitations
Manual state weighting and updating as per RicardoSantos's Function Markov Process is required:
What's being plotted for are the first three layers, but without the full range of the input_dimensions, hidden_dimensions:
syn_0 (blue)
syn_1 (green)
syn_h (red)
While there's more than a few technical hurdles to overcome (i.e. potential pine issues from max variables to runtime/compile limits, no real arrays, functions to do state updates RichardoSantos Markov Function style, etc), I'm fairly confident a limited working model should be possible to create in Pine.
Indicators Input Window Length - Problems And SolutionsIntroduction
Most technical indicators possesses a user defined input window length, this input affect the indicator output and for a long time, have been the cause of many innovations in technical analysis.
In this post i want to discuss the effects and particularities of indicators inputs window length, the challenges they introduce in trading and their effect when paired with machine learning forecasting models, i hope this post will be easy to read, let me know if you had difficulties understanding it.
Speed And Efficiency Problems
An input window length can involve the number of data processed by the indicator, therefore higher window length's will process more data, which result in a slower computation time, therefore in high-frequency/algorithmic trading where response time matter, maximizing the profitability might be made at the cost of the indicator response time, and even if computerized trading has been praised for its speed, small lag times can actually affect your strategy, therefore one might enter a trade at a different value than the targeted price.
Note : High frequency trading (hft) is a commonly mistaken term, one might believe that hft require the trader to open and close a large number of trades in a short amount of time, in reality hft is related to the "rate at which data is processed".
Solution - Efficient Data Processing
Solutions have been proposed in order to make certain tools more efficient. For example the simple moving average is a common tool that is the basis of many other indicators, its calculation involve summing the length last data points and diving this sum by length . In signal processing, such tool require what is called "memory", the data points must be stored in order for them to be processed, this is extremely inefficient and slow, therefore alternatives have been proposed, one of them is still mainly used in technical analysis today and is called the exponential moving average (ema), the process of computing an exponential moving average is called exponential averaging, and has the form of :
ema = sc*input+(1-sc)*(past ema value)
where sc is called the smoothing constant where 1 > sc > 0 . We only need 2 data values in order to perform this computation, lets denote a moving average of period length sma(input,length) , we can estimate it using exponential averaging with sc = 2/(length+1) . The computation time of the exponential moving average is way lower than the one of the sma . This is the most elegant and efficient estimation of the simple moving average.
The exponential moving average is the simplest "IIR Filter", or infinite response filter, those filters are as well extremely efficient since they use recursion. Exponential averaging is also the core of many adaptive indicators. In my experience, recursion will always let you create extremely efficient tools.
Window Length And Optimization Problems
Optimization is a branch of mathematics that help us find the best parameters in order to maximize/minimize a certain function, and thanks to computers this process can be made faster. Optimizing technical indicators during backtesting involve finding the input window length (set of inputs if there are more than 1 input) that maximize the profit of a strategy.
The most common approach is brute forcing, in which we test every indicator inputs window length combination and keep the one that yield the best results. However optimization is still computationally intensive, having 2 indicators already involve a high number of combinations. This is why it is important to select a low number of indicators for your strategy. But then other problems arise, the best input window length (set of inputs) might change in the future. This is due to the fact the market price is non-stationary and one of the reasons technical indicators are looked down.
In order to deal with this problem, we can propose the following solutions :
Use indicators/Information with no input window length -> Vwap/Volume/True Range/Cumulative Mean...etc.
Study the relationship between the optimal input window length and price evolution -> Regression analysis
Forecast the optimal input window length -> Forecasting
The last two are extremely inefficient, kinda nightmarish, and would be time consuming if one use a serious backtesting procedure. However the first solution is still appealing and might actually provide a efficient result.
Machine Learning Forecasting - Performance And Technical indicators Input Window Length Dependency
Technical indicators outputs can be used as inputs for machine learning algorithms. We could think that we also need to optimize the input window length of the indicators when using machine learning (which would lead to high computations time, machine learning already involve optimization of a high number of parameters), however a research paper named "Forecasting price movements using technical indicators: Investigating the impact of varying input window length" by Yauheniya Shynkevicha, T.M. McGinnitya, Sonya A. Colemana, Ammar Belatrechec and Yuhua Li highlight an interesting phenomenon, the abstract tell us that :
"The highest prediction performance is observed when the input window length is approximately equal to the forecast horizon"
In short, if you want to forecast market price 14 step ahead with a machine learning model, you should use indicators with input window length approximately equal to 14 as inputs for the model in order to get the best performance. This would allow to skip a lot of optimizations processes regarding the technical indicators used in the model. They used 3 different type of ML algorithms, support vector machine (svm) , adversarial neural networks (ann) and k nearest neighboring (knn) , which reinforce their conclusion.
In the paper, we can see something interesting with the indicators they selected as inputs, they used : A simple moving average, an exponential moving average, the average true range, the Average Directional Movement Index, CCI, ROC, RSI, %R, stochastic oscillator.
First thing we can see is that they used the exponential moving average instead of the wilder moving average for certain calculations, which i think is a good choice. We can also see they used many indicators outputting the same kind of information, in this case we often talk about "Multicollinearity", for example :
The CCI, ROC, RSI, %R, Stochastic output similar information, all remove the trend in the price, the CCI and ROC are both centered around 0 and the %R, RSI and stochastic oscillator around 50. The SMA and EMA also output similar information.
In technical analysis this practice is often discouraged since the indicators will output the same kind of information, this lead to redundancy. However such practice has been seen a lot in machine learning models using technical indicators. Maybe that a higher amount of multicollinearity between indicators allow to strengthen the relationship between the forecast horizon and the indicators input window length.
Conclusion
We talked (a lot) about indicator inputs window length, what problems they cause us and how we can find solutions to those problems. Also we have seen that the forecasting performance of ML models can be higher when they use indicators outputs with input window length equal to the forecasting horizon. This can make to make the process of forecasting financial market price with ML models using technical indicators more efficient.
ML is a recurring subject in financial forecasting, those algorithms offer the hope to make technical indicators more useful, and indeed, technical indicators and ML models can benefits from each others, however it is sad to observe that classical indicators are mainly used instead of newer ones, but its also encouraging in the sense that more research can be done, using newer material/procedures.
Thanks for reading !
Neural Network Analysis 2009-2018 USDTRY Next Week Prediction01-Jan-2009 - 23-Oct-2018 Daily Close price data neural network analysis(Nonlinear autoregressive - NAR 1-2 delays, 10 neurons- 1 hidden layer, Levenberg-Marquardt backpropagation)
5.7420 23.10.2018
5.8501 24.10.2018
5.9547 25.10.2018
6.0453 26.10.2018
6.1165 27.10.2018
6.1667 28.10.2018
6.1978 29.10.2018
6.2138 30.10.2018
6.2195 31.10.2018
Wild A.I. Forecast & TA Say_________?Bitcoin/USD Daily Bars, LOG Scale, 06/12/18, 11:10 PM EST, by Mike Mansfield
(REDONE SO CYCLES FIT)
Hi trader friends, this is one of our most detailed and important Bitcoin updates. Our first post with screenshots of our deep learning A.I. project’s 12 month forecast. I will post those screenshots after this clears the 15 minute publishing hold. So let’s get to it.
SUMMARY:
BTCUSD is likely to drop another 30-60% over the next 4-12 months, possibly 18 months.
Short-term cycle lows are expected in July and October 2018.
The mid-July 2018 cycle low should be followed by a 3-8 week trading bounce. Then, another big down wave into early to mid October 2018.
October could be “the low,” but we are thinking next year, based on our data.
The final cycle low, or momentum lows are expected to be in February 2019 (A.I. low date), or June 9th, 2019 (pink cycle low), or possibly January 2020 (blue cycle low).
Initial price targets lower are $4935, then $2967-2380 (or lower).
A.I. project forecasts 12 month low @ $2774, due in February 2019.
The deep learning A.I. project’s forecast is confirmed by Elliott Waves, Andrews Lines, Wolfe Waves, and a few other techniques shown here, most suggesting much lower prices ahead.
Bitcoin has broken its “low to low” support trend-line.
Will likely close below the Andrews Median Line (orange pitchfork's mid line), which, would put Bitcoin below the black dotted trend-line drawn off the lowest closing prices of 2018. Once broken, the first three price targets to the downside become quite likely.
Time Cycle Caveat:
We are working on pattern matching comparison called an analog. This suggests an October 2018 low. That fits the secondary A.I model's forecast, but not the top performing algorithm's forecast (February 2019 low). Remember, two of the longer-term cycles are pointing to either a June 2019 low (pink cycle), or January 2020 low (blue cycle). Therefore, the October low this year seems less likely thus far, but not out of the question.
Truth Seekers & The Crypto Land Bear:
Are you a “truth seeker?" Really a truth seeker? If you want to be a more objective trader, be a truth seeker. Try to be a truth seeker in all things. Accept that there are many possibilities in life, as in trading. Seek to find the ultimate “truth," but while doing it, at least for trading, go with the highest probabilities. Else, you'll live in La La Land of wannabe moonshots for every crypto market. Of course, some tokens moonshot, even in a bear market. But, as the top 10-25 go, so goes most of the market.
PRICE TARGETS:
Target 1: $4935, likely minimum target. It's the prior Wave (3) high on Sept 1st, 2017.
The “forever trend-line” (black upward sloping) is currently around $3700. Yet, by the time the market could logically get there, that trend-line would likely be around $4450-$4950.
Target 2: $2972, the June 2017 high of the prior Wave V of Wave (3). This is close to our
A.I. derived target. Outcome, likely.
Target 3: $2774, A.I. deep learning 12 month projected low, due February 2019.
Since target 2 and 3 are so close, just a blip in Crypto Land, I would be targeting a low between those two figures, unless more data changes the A.I. forecasts. Target, likely.
Target 4: $2380, September 15th, 2017, Wave ii low.
Target 5: $1138, where the Wolfe Wave support (purple line) and the prior Wave ( III) high of large degree meet. This is a potential attractor point (less likely now). There is a far lower target, but I'll let you find it on the chart.
----------
The good news for bulls is that the Alternative Wave Count (lighter blue) could very well be correct. That would mean one more all-time high to complete Bitcoin’s Wave (V) advance.
BOTTOM LINE:
Down 30-60%, Feb. or June low, around $2774.
DISCLOSURE:
This analysis is meant for educational purposes only. You trade at your own risk!
Cheers and best wishes,
Michael Mans
BTC/USD (Bitcoin) Wild A.I. Forecast & TA Say_________?Bitcoin/USD Daily Bars, LOG Scale, 06/12/18, 9:00 PM EST, by Mike Mansfield
Hi trader friends, this is one of our most detailed and important Bitcoin updates. Our first post with screenshots of our deep learning A.I. project’s 12 month forecast. I will post those screenshots immediately after this clears the 15 minute publishing hold. So let’s get to it.
BOTTOM LINE:
BTCUSD is likely to drop another 30-60% over the next 4-12 months, possibly 18 months.
Short-term cycle lows are expected in July and October 2018.
The mid-July 2018 cycle low should be followed by a 3-8 week trading bounce. Then, another big down wave into early to mid October 2018.
October could be “the low,” but we are thinking next year, based on our data.
The final cycle low, or momentum lows are expected to be in February 2019 (A.I. low date), or June 9th, 2019 (pink cycle low), or possibly January 2020 (blue cycle low).
Initial price targets lower are $4935, then $2967-2380 (or lower).
A.I. project’s 12 month low price forecast's is $2774, due in February 2019.
The deep learning A.I. project’s forecast is also confirmed by Elliott Waves, Andrews Lines, Wolfe Waves, and a few other techniques shown here, most suggesting much lower prices ahead.
Bitcoin has broken its “low to low” support trend-line. It also appears to on its way to closing below the Andrews Median Line (orange pitchfork's mid line), which, would put Bitcoin below the black dotted trend-line drawn off the lowest closing prices of 2018. Once broken, the first three price targets to the downside become quite likely.
Time Cycle Caveat:
We are working on pattern matching comparison called an analog. This also suggests an October 2018 low. That fits the secondary A.I model's forecast, but not the top performing algorithm forecast that calls for a February 2019 low. And remember, two of the longer-term cycles are pointing to either a June 2019 low (pink cycle), or January 2020 low (blue cycle). Thus, the October low this year seems less likely thus far, but not out of the question.
Truth Seekers & The Crypto Land Bear:
Are you a “truth seeker?" Really a truth seeker? If you want to be a more objective trader, be a truth seeker. Try to be a truth seeker in all things. Accept that there are many possibilities in life and in trading. Seek to find the ultimate “truth," but while doing it, at least for trading, go with the highest probabilities. Else, you'll live in La La Land of wannabe moonshots for every crypto market. Of course, some tokens will moonshot, even in a bear market, but as the top 10-25 go, so goes most of the market.
PRICE TARGETS:
Target 1: $4935, likely minimum target. It's the prior Wave (3) high on Sept 1st, 2017.
The “forever trend-line” (black upward sloping) is currently around $3700, but by the time the market could logically get there, that trend-line would likely be around $4450-$4950.
Target 2: $2972, the June 2017 high of the prior Wave V of Wave (3). This is close to our A.I. derived target. Likely.
Target 3: $2774, A.I. deep learning 12 month projected low for February 2019. Since target 2 and 3 are so close, just a blip in Crypto Land, I would be targeting a low between those two figures, unless more data changes the A.I. forecasts. Target, likely.
Target 4: $2380, September 15th, 2017, Wave ii low.
Target 5: $1138, where the Wolfe Wave support (purple line) and the prior Wave (III) high of large degree meet. This is a potential attractor point (less likely now). There is a far lower target, but I'll let you find it on the chart.
----------
The good news for bulls is that the Alternative Wave Count could very well be correct. That would mean one more all-time high to complete Bitcoin’s Wave (V) advance.
Summary:
Down 30-60%, Feb. or June low around $2774.
DISCLOSURE:
This analysis is meant for educational purposes only. You trade at your own risk!
Cheers and best wishes,
Michael Mansfield CIO
Heffae USDJPY - Cloud interaction & predictive path fittingHeffae Clouds functioning as adaptive support and resistance in FOREX markets. High validity path fitting, cloud color provides granular detail on path validity, constitution of paths use default maths with a higher offset.
This demonstrates the flexibility and power of Heffae Clouds. The ability to automatically adapt and produce high validity path-fitting for many different asset classes.
TSLA 10-Day Ahead Prediction 05/21 - 06/04 PeriodArtificial Intelligence/Deep Learning Enabled 10 Day Ahead Predicted values for Tesla Inc ( TSLA ) have been plotted on the chart.
You can see FinBrain's previous prediction for TSLA stock (04/30 - 05/11 Period) on the same chart as well.
The method used in this prediction is Deep Learning/Artificial Neural Network based, and using complex mathematical models/methodologies to analyze and extract hidden time series features in vast amounts of TSLA related data. The "trained" and optimized neural network then generates the multi-day ahead predictions which are plotted on the chart and given in the explanation below.
The expected 5 Day Change is 2.851 %
The expected 10 Day Change is 5.356 %
Predictability Indicator is calculated as : 0.726
Predicted 10-Day Ahead Prices are as follows:
Mon May 21 Tue May 22 Wed May 23 Thu May 24 Fri May 25
274.92 275.85 277.88 281.06 284.71
Tue May 29 Wed May 30 Thu May 31 Fri Jun 01 Mon Jun 04
287.11 288.48 289.88 291.06 291.64
Please note that outliers/non-linearities might occur, however FinBrain's Artificial Intelligence/Deep Learning Enabled predictions indicate the softened/smoothed moving direction of the stocks/commodities/World Indices/ETFs/Foreign Currencies/Cryptocurrencies.