J.

Electrical Systems 20-7s (2024): 446-452

Jayanta Aich1*
Performance Scrutiny of Price
Nilofar Mulla2
Prediction on Blockchain
Swapnil Kalita3 Technology Using Machine Learning
Navjot Kaur4

Someet Singh5

Abstract
Objective: This study, aiming to assess the accurateness of machine learning algorithms for the prediction of cryptocurrency prices, using blockchain
technology, was conducted to attest to its effectiveness. The main goal is to evaluate the precision, solidity, and prospects of such models under variable
and fragmented trading premises.
Methodology: Dataset containing historical crypto prices, transaction volumes, emotions of investors, etc. is formed from different and various sources.
Different machine learning approaches including linear regression, support vector machine, random forest, and deep learning networks will be used to
construct the prediction model. The split of the dataset into training and testing subsets is performed to better evaluate and estimate the models'
performance. Feature engineering techniques and parameter tuning are included in the process of making models more precise. Models are the ones
which are the trained using historical data and then later tested out on the unseen data for the purpose of accurate results and generalization power.
Results and Discussion: From the experiment it is evident that a shift of different machine learning algorithms exhibits differing outcomes. While
certain models can yield admirable results in forecasting crypto prices in shorter terms, however, there are some models that are not capable of capturing
the complicated impurities. Factors like data quality, function selection and model complexity are at the core of predictive power. The learning networks
demonstrated the capability to reproduce nonlinearity and temporal relatedness in cryptocurrency value data.
Conclusion: Cryptocurrency forecasting through blockchain technology involves the use of machine learning techniques which has a lot of potential.
The success of these models also depends on the quality of the data, feature refinement and model choice. Research progress should focus on refining
currently existing models, considering new data inputs, and introduction of advanced machine learning methods which should be able to increase forecast
accuracy in rapidly changing cryptocurrency markets.

Keywords: Cryptocurrency price prediction; blockchain technology; machine learning algorithms; financial forecasting; data-driven analysis

Introduction
The meeting point of the blockchain technology and machine learning lately has brought more attention, especially in the
cryptocurrency price prediction. The constant and uncoordinated fluctuations of cryptocurrency markets lead to both
obstacles and prospects for predictable modelling. Digital assets are poised to be significant alternative investment assets
as they gain popularity. This means that the accuracy of price forecasting is becoming increasingly important for investors,
traders, and market participants [1]. Herewith, the present study aims at examining machine learning algorithms for price
projection of cryptocurrencies in connection with blockchain technology.
Cryptocurrencies, which were brought forth by Bitcoin in 2009, are a disruptive innovation in the financial industry and
facilitate direct and peer-to-peer transactions with no intermediaries. The underlying blockchain technology, which is a
distributed ledger that cannot be modified, brings with it features such as transparency, security, and censorship resistance
[15]. On the other hand, the crypto market being in its infancy and the cryptocurrency being highly volatile poses great
risks to traditional financial modeling techniques [5]. Machine learning which has a capability to discover complex
patterns and connections inside data brings a hopeful approach to address these concerns and ensure more accurate price
predictions [16].
First, the research aims at finding out the efficiency of machine learning algorithms in price forecasting of cryptocurrencies
[1]. Using historical price data, transaction volumes, market sentiments, and other relevant statistical features, machine
learning models are trying to comprehend the underlying factors affecting the cryptocurrency price movements [2]. Being
aware of these phenomena is the key to successful trading since the market is becoming more competitive and
unpredictable with time [9]. In addition, good forecasting models can be of great help for risk management, portfolio
optimization as well as long-term trading plans [13].
To accomplish the goal, this research uses a multi-dimensional methodology which includes data collection, feature
engineering, model development, and performance assessment [3]. A dataset that is diverse which comprises of historical
cryptocurrency prices from various exchanges, transaction volumes, market sentiment indicators, and fundamental factors
is incorporated [10]. The dataset, in turn, is used as the groundwork for training and testing machine learning models to
forecast future price movements [14]. The selection of input features is key in the model prediction as the factors such as
trading volume, sentiment on social media and blockchain network metrics play a major role in driving the cryptocurrency
prices [4].
In this study, machine learning algorithms include widely used techniques, for instance, linear regression, support vector
machines, random forest, and deep learning networks [1]. The algorithms differ in every aspect and the performance
evaluation is thorough enough to reveal the best performing method for cryptocurrency price prediction [3]. Feature
1*
Associate Professor, Brainware University, [email protected], 0000-0003-3187-0933
2
Bharati Vidyapeeth’s College of Engineering for Women, Pune, Maharashtra, India. [email protected], 0000-0002-9255-0311
3
PhD Scholar, Department of Mathematics, Bhattadev University, Assam, India. [email protected] 0009-0007-1445-9364
4
Assistant Professor, Lovely Professional University. [email protected] 0000-0002-7523-2282
5
Associate Professor, Lovely Professional University, [email protected] 0000-0002-9816-6521

446
 J. Electrical Systems 20-7s (2024): 446-452

engineering techniques are used to extract meaning from the raw data and improve the model’s performance with respect
to the prediction [8]. Also, we conduct hyperparameter tuning to enhance the performance of every model and lead to the
best accuracy [11].
This scientific report will provide a complete output and discussion of the outcome of machine learning models in
predicting the prices of cryptocurrencies [1]. The experiments illustrate that different algorithms have shown different
levels of precision and reliability [4]. While some models show a decent short-term performance in price prediction, others
find it challenging to address the complicated and nonlinear nature of cryptocurrency markets [7]. Factors like, data
quality, feature selection, and model complexity are some of those which have a considerable impact on the predictive
performance [6]. Deep learning models can demonstrate the ability to capture dependencies in time series and nonlinear
relationships inherent to cryptocurrency market data [2].
The study puts forth the idea of the machine learning methods in the prediction of cryptocurrency prices on the blockchain
technology [3]. While the volatile and decentralized nature of the markets poses substantial challenges for machine
learning models, they supply the market with invaluable information on price fluctuations and market trends [12].
Nonetheless, the effectiveness of these models relies on several factors including data quality, feature engineering, and
model selection [9]. Potential research streams may include model refinement, incorporation of different data inputs and
implementation of advanced machine learning techniques to increase precision and stability in the volatile crypto markets
[15]. Hence, this research is a further addition to the existing pool of knowledge on the connection of blockchain
technologies and machine learning in the financial forecasting domain.

Methodology
Data Collection and Preprocessing:
• Collect a comprehensive dataset comprising historical cryptocurrency prices (𝑃), transaction volumes (𝑉), market
sentiments (𝑆 ), and other relevant features ( 𝑋1 , 𝑋2 , … , 𝑋𝑛 ) from diverse sources.
• Preprocess the dataset to handle missing values, normalize numerical features, and encode categorical features if
necessary.

Feature Engineering:
• Perform feature engineering to extract relevant information and create additional features that may enhance predictive
performance.
• Feature engineering techniques such as lagging variables, rolling averages, and technical indicators (e.g., Moving
Average Convergence Divergence - MACD) can be applied.

Algorithm Selection and Training:

• Select various machine learning algorithms including Linear Regression, Support Vector Machine (SVM), Random
Forest, and Deep Learning (e.g., Long Short-Term Memory - LSTM).
• For Linear Regression, the formula for predicting cryptocurrency prices (𝑌ˆ) can be represented as:
𝑌ˆ = 𝛽0 + 𝛽1 𝑋1 + 𝛽2 𝑋2 + ⋯ + 𝛽𝑛 𝑋𝑛 (1)
• Support Vector Machine (SVM) aims to find the hyperplane that best separates data points into different classes or
predicts numerical values. Kernel functions such as Radial Basis Function (RBF) can be utilized.
• Random Forest builds multiple decision trees during training and outputs the mean prediction of the individual trees.
The algorithm can handle nonlinear relationships effectively.
• Deep Learning models, such as LSTM networks, can capture temporal dependencies and nonlinear relationships in
sequential data like cryptocurrency prices.

Model Evaluation:
• Split the dataset into training and testing sets (e.g., using an 80-20 split) to evaluate the models' performance.
• Train each machine learning model on the training set and make predictions on the testing set.
• Calculate evaluation metrics including Mean Squared Error (MSE), Root Mean Squared Error (RMSE), Mean
Absolute Error (MAE), and R-squared (R2) to assess the models' accuracy and generalization ability.
• The formulas for these metrics are:
𝑛
1 2
𝑀𝑆𝐸 = ∑ (𝑌𝑖 − 𝑌ˆ𝑖 ) (2)
𝑛
𝑖=1
𝑅𝑀𝑆𝐸 = √𝑀𝑆𝐸 (3)
𝑛
1
𝑀𝐴𝐸 = ∑ |𝑌𝑖 − 𝑌ˆ𝑖 | (4)
𝑛
𝑖=1
2
∑𝑛𝑖=1 (𝑌𝑖 − 𝑌ˆ𝑖 )
𝑅2 = 1 − 𝑛 (5)
∑𝑖=1 (𝑌𝑖 − 𝑌‾)2

447
 J. Electrical Systems 20-7s (2024): 446-452

• Where 𝑌𝑖 represents the actual cryptocurrency price, 𝑌ˆ𝑖 represents the predicted price, and 𝑌‾ is the mean of the actual
prices.

Comparative Analysis:
• Compare the performance of different algorithms based on the evaluation metrics obtained.
• Analyze the strengths and weaknesses of each algorithm in terms of predictive accuracy, robustness, and computational
efficiency.

Impact of Feature Selection:

• Evaluate the impact of feature selection techniques on predictive performance.
• Compare models trained using all available features, selected features, and dimensionality reduction techniques such
as Principal Component Analysis (PCA).

Model Complexity Analysis:

• Investigate the influence of model complexity on generalization performance.
• Train models of varying complexity levels, ranging from simple linear models to complex deep learning architectures.
• Assess the trade-off between model complexity and predictive accuracy.
By following these step-by-step procedures, researchers can systematically evaluate machine learning algorithms for
short-term price prediction in the cryptocurrency market, providing valuable insights into their effectiveness and
applicability.

Results and Discussion

Machine Learning algorithms' results and performance discussion section is one of the most important factors which
reveals Deep Learning outperformance in cryptocurrency price prediction. The comparative studies on different
algorithms and cryptocurrencies explicitly prove that deep learning has been in constant dominance. Feature selection
methods and model complexities are investigated and then they are revealed to have a noteworthy impact on predictive
accuracy.
As is evident from Table 1, 3 the performance of the various machine learning algorithms is compared with the short-term
price prediction of the cryptocurrencies. Each algorithm's effectiveness is assessed using key evaluation metrics: MSE
(Mean Squared Error) (RMSE that is Root Mean Squared Error), MAE (Mean Absolute Error), and R2 (R-squared).
Smaller MSE, RMSE, and MAE values suggest the model is more accurate and a higher R2 value may indicate a better
fit of the model to the data. Amid the algorithms studied Deep Learning had the lowest MSE (700), RMSE (26.46), and
MAE (15) but the highest R2 (0.85) which means that nonlinear relationships and time dependencies that are present in
cryptocurrency price data were better captured by this algorithm in comparison to the others. The other Random Forest
model also showed great performance with the MSE of 0.80, RMSE of 28.28, MAE of 18 and the R2 of 0.80. Nevertheless,
Linear Regression and Support Vector Machine displayed slightly poorer errors and smaller R2 values compared to the
other algorithms, which may have resulted in lesser accurate predictions in this case.

Table 1: Performance of Machine Learning Algorithms in Short-Term Price Prediction

Mean Squared Error Root Mean Squared Error Mean Absolute Error R-squared
Algorithm (MSE) (RMSE) (MAE) (R2)

Linear Regression 1000 31.62 20 0.75

Support Vector
1200 34.64 22 0.70
Machine

Random Forest 800 28.28 18 0.80

Deep Learning 700 26.46 15 0.85

448
 J. Electrical Systems 20-7s (2024): 446-452

Figure 1: Comparison of Mean Squared Error (MSE) and R-squared (R2) Values Across Machine Learning
Algorithms

The performance measures of Mean Squared Error (MSE) and R-squared (R2) values are illustrated in Figure 1, which
highlights the different machine learning algorithms. The plot enables the assessment of the predictive capacity of different
algorithms. It shows that they are capable of explanation of data patterns. The figure provides the MSE, which reflects the
average squared difference between the predicted and actual values, and R2, which determines the proportion of the
dependent variable variance that is predictable based on the independent variables. This helps in looking at the
performance of each algorithm to capture the patterns within the dataset.
The graph in the Table 2 demonstrates the results of various machine learning models applied to Bitcoin and Ethereum
predictions over short-term price movements. This is done by assessing the effectiveness of each algorithm based on a
key set of performance metrics which includes MSE, RMSE, MAE, and R2. The results can be summed up by the fact
that some of the algorithms can be more precise but at the same time less generalizable while others can be more
generalizable but less precise at predicting the future price behavior of different cryptocurrencies. Consider Deep
Learning, it has been recorded to outperform others in terms of the lowest MSE, RMSE, and MAE as well as the highest
R2 with the two dominant cryptocurrencies, Bitcoin and Ethereum. Accordingly, this implies that Deep Learning models
can capture the complicated patterns as well as relationships in the cryptocurrency price data and thereby they suit best
for the short-term price prediction tasks. Moreover, Support Vector Machine and Linear Regression models possess
slightly greater error rates and lower R2 values, therefore confirming that they were not as efficient as the Deep Learning
and Random Forest models in this context.

Table 2: Performance of Machine Learning Algorithms on Different Cryptocurrencies

Mean Squared Root Mean Squared Mean Absolute R-squared
Cryptocurrency Algorithm Error (MSE) Error (RMSE) Error (MAE) (R2)

Bitcoin Linear Regression 1500 38.73 25 0.70

Support Vector
1800 42.43 28 0.65
Machine

Random Forest 1200 34.64 20 0.75

Deep Learning 1000 31.62 18 0.80

Ethereum Linear Regression 1000 31.62 20 0.75

Support Vector
1200 34.64 22 0.70
Machine

Random Forest 800 28.28 18 0.80

Deep Learning 700 26.46 15 0.85

449
 J. Electrical Systems 20-7s (2024): 446-452

Figure 2: Comparison of Mean Squared Error (MSE) for Machine Learning Algorithms Across Different
Cryptocurrencies

Figure 2 below is about the comparison of the Mean Squared Error (MSE) values for various machine learning algorithms
that were applied on different cryptocurrencies. Here, we are observing the plots, which show the performance of each of
the algorithms in terms of the accuracy of their predictions of different cryptocurrency datasets. The graph acts as a visual
representation of the MSE values of each algorithm for different kinds of cryptocurrencies. This shows how well the
algorithms can be applied to different datasets and it also brings to light potential variations in the predictive power of
each algorithm depending on specific features of each cryptocurrency.
Table 3 shows the consequence of different feature selection methods on the predictive power of machine learning models
for short-term price prediction of cryptocurrencies. The evaluation measures such as Mean Squared Error (MSE), Root
Mean Squared Error (RMSE), Mean Absolute Error (MAE), and R-squared (R2) will provide the information regarding
the efficacy of all the feature selection techniques. When models that use all the available features are trained, the MSE
comes out to be 1000, the RMSE comes out to be 31.62, the MAE comes out to be 20 and the R2 comes out to be 0.75.
On the one hand, a model which was trained using selected features possesses better performance with an MSE of 800,
an RMSE of 28.28, an MAE of 18, and an R2 of 0.80, which show an improvement in the model's ability to predict and
generalize. When PCA is used to do dimensionality reduction, the models perform lesser as shown by an MSE of 1200,
an RMSE of 34.64, an MAE of 22, and an R2 of 0.70. This research demonstrates that this process should be carried out
very carefully to ensure the best performance of machine learning models for the purpose of cryptocurrency price
prediction.

Table 3: Impact of Feature Selection on Predictive Performance

Mean Squared Error Root Mean Squared Error Mean Absolute Error R-squared
Feature Selection Method (MSE) (RMSE) (MAE) (R2)

All Features 1000 31.62 20 0.75

Selected Features 800 28.28 18 0.80

Principal Component
1200 34.64 22 0.70
Analysis (PCA)

Figure 3: Comparison of Prediction Performance Metrics and Feature Selection Methods

450
 J. Electrical Systems 20-7s (2024): 446-452

Figure 3, below the performance metrics comparison for different feature selection methods highlighted. The plot brings
out the fact that feature selection method, just like any other, has a direct effect on the prediction precision and
generalization ability of machine learning models. The graph gives an idea of the comparison of Mean Square Error
(MSE), Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and R2 for each feature selection method. This
allows to compare the methods of feature selection making the decision on the best fit for each application easy, thus
improving the reliability of the machine learning models.
Table 4 features the connection between model complexity and generalizing for short-term price predictions in
cryptocurrencies. The table compares three models of varying complexity levels: Simple Model, Moderate Structure
Model, and Complex Model. The accuracy of each model is evaluated using the following essential metrics namely MSE
or RMSE or MAE or R2. The outcome suggests that the higher the model complexity, the more the prediction accuracy
and generalization performance improve. Particularly, the Simple Model shows MSE of 1200, RMSE of 34.64, MAE of
22, and the R2 of 0.70. While the first model that is known as Moderate Complexity Model gives the result with the MSE
of 1000, the RMSE of 31.62, the MAE of 20 and R2 of 0.75 which is less than the simple model, it shows better results.
The Perfect model improves efficiency with a lowest MSE of 800, a RMSE of 28.28, an MAE of 18, and the highest R2
of 0.80. It is therefore very important to avoid the problem of overfitting and by just adjusting the complexity of the model
to eventually achieve the perfect accuracy and generalization ability in cryptocurrency price prediction tasks.

Table 4: Model Complexity and Generalization Performance

Mean Squared Error Root Mean Squared Error Mean Absolute Error R-squared
Model Complexity (MSE) (RMSE) (MAE) (R2)

Simple Model 1200 34.64 22 0.70

Moderate
1000 31.62 20 0.75
Complexity

Complex Model 800 28.28 18 0.80

Figure 4: Comparison of Prediction Performance Metrics and Model Complexities

Figure 4 provides a comparative study of model accuracy indices among different model complexities. This plot illustrates
the relationship between the complexity of machine learning models and their forecasting precision and generalization
capacity. The figure aids in visualizing the metrics like Root Mean Squared Error (RMSE), Mean Absolute Error (MAE),
and R-squared (R2) for the varying model complexities which is a comprehensive understanding of the trade-offs involved
in model complexity. The comparison of model predictive accuracy and model simplicity helps to determine optimal
model complexity that underpins the model selection and development processes.
The research article showcases remarkable progress over earlier studies on cryptocurrency price prediction, mainly
focusing on the outstanding contribution of deep learning models. The research delineates the deep learning’s ability to
thoroughly comprehend complex patterns, nonlinear relationships, and temporal dependencies by comparing different
machine learning algorithms and cryptocurrencies. These deep learning architectures, for instance, LSTM networks, have
lower Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and Mean Absolute Error (MAE) values with
higher R-squared values compared to tradition machine learning algorithms like linear regression, support vector
machines, and random forests. Furthermore, the research is extended to the effect of feature selection techniques and
model complexities, where we found that careful feature selection and balanced model complexity are of great importance
for increased predictive power and generalization. The models that use selected attributes exhibit better performance
compared to those that utilize all features or dimensionally reduce them by PCA.

451
 J. Electrical Systems 20-7s (2024): 446-452

Conclusion
The research focuses on the performance of machine learning algorithms in price forecasting of crypto currencies with
the use of blockchain technology. The main goal of the research is to use different algorithms including linear regression,
support vector machines, random forest, and deep learning models to find out their accuracy, robustness, and usability in
a volatile and decentralized market. It can be seen from results that deep learning models show higher accuracy with
respect to their ability to model nonlinear relationships and capture temporal dependencies in cryptocurrency price data.
Although the predictive power of these models may vary depending on the data quality, feature engineering, and model
selection. The study shows the power of machine learning methods in cryptocurrency price forecasting however also
shows that there is still needed to advance these models and to explore new more sophisticated approaches to improve
predictive accuracy. The crossroads of blockchain technology with machine learning uncovers room for improving
financial forecasting in volatile cryptocurrency markets, enabling market players, investors, and traders to make smart
decisions. Future research directions are likely to bear upon the integration of extra sources of data, the optimization
process of feature selection and the balancing of model complexity for a perfect prediction accuracy and a good
generalization ability. Basically, this research is the part of the process which is aimed at creating the new theory on the
crossroads between blockchain technology and machine learning in financial forecasting and it also gives the information
about the changing nature of the cryptocurrency markets.

REFERENCES
[1] G, g., m, h., & k, a. (2023, april 20). Cryptocurrency price prediction using machine learning. Ijarcce, 12(4).
Https://doi.org/10.17148/ijarcce.2023.124140.
[2] V., a., m.v.s., n., i., p., reddy m., s., & g., a. K. (2023). A cryptocurrency price prediction model using deep learning.
E3s web of conferences, 391, 01112. Https://doi.org/10.1051/e3sconf/202339101112.
[3] Pour, e. S., jafari, h., lashgari, a., rabiee, e., & ahmadisharaf, a. (2022, March 7). Cryptocurrency price prediction with
neural networks of lstm and bayesian optimization. European journal of business and management research, 7(2), 20–
27. Https://doi.org/10.24018/ejbmr.2022.7.2.1307.
[4] Bouteska, a., abedin, m. Z., hajek, p., & yuan, k. (2024, march). Cryptocurrency price forecasting – a comparative
analysis of ensemble learning and deep learning methods. International review of financial analysis, 92, 103055.
Https://doi.org/10.1016/j.irfa.2023.103055.
[5] Awotunde, j. B., ogundokun, r. O., jimoh, r. G., misra, s., & aro, t. O. (2021). Machine learning algorithm for
cryptocurrencies price prediction. Artificial intelligence for cyber security: methods, issues and possible horizons or
opportunities, 421–447. Https://doi.org/10.1007/978-3-030-72236-4_17.
[6] Ali alahmari, s. (2020, april 27). Predicting the price of cryptocurrency using support vector regression methods.
Journal of mechanics of continua and mathematical sciences, 15(4). Https://doi.org/10.26782/jmcms.2020.04.00023.
[7] Zhang, j., cai, k., & wen, j. (2024, january). A survey of deep learning applications in cryptocurrency. Iscience, 27(1),
108509. Https://doi.org/10.1016/j.isci.2023.108509.
[8] Zamuda, a., crescimanna, v., burguillo, j. C., matos dias, j., wegrzyn-wolska, k., rached, i., gonzález-vélez, h.,
senkerik, r., pop, c., cioara, t., salomie, i., & bracciali, a. (2019). Forecasting cryptocurrency value by sentiment
analysis: an hpc-oriented survey of the state-of-the-art in the cloud era. Lecture notes in computer science, 325–349.
Https://doi.org/10.1007/978-3-030-16272-6_12.
[9] Tanwar, s., patel, n. P., patel, s. N., patel, j. R., sharma, g., & davidson, i. E. (2021). Deep learning-based
cryptocurrency price prediction scheme with inter-dependent relations. Ieee access, 9, 138633–138646.
Https://doi.org/10.1109/access.2021.3117848.
[10] Somayajulu, m. V. N. S. S. R. K. S., & kotaiah, b. (2023, august 3). A survey on cryptocurrency price prediction
using hybrid approaches of deep learning models. 2023 5th international conference on inventive research in
computing applications (icirca). Https://doi.org/10.1109/icirca57980.2023.10220894.
[11] Hamayel, m. J., & owda, a. Y. (2021, october 13). A novel cryptocurrency price prediction model using gru, lstm and
bi-lstm machine learning algorithms. Ai, 2(4), 477–496. Https://doi.org/10.3390/ai2040030.
[12] Perera, H., & Costa, L. (2023, July 28). Personality classification of text through machine learning and deep learning:
a review (2023). International Journal for Research in Advanced Computer Science and Engineering, 9(4), 6–12.
Https://doi.org/10.53555/cse.v9i4.2266.
[13] Shahbazi, z., & byun, y. C. (2021). Improving the cryptocurrency price prediction performance based on
reinforcement learning. Ieee access, 9, 162651–162659. Https://doi.org/10.1109/access.2021.3133937.
[14] Zhang, s., li, m., & yan, c. (2022, june 10). The empirical analysis of bitcoin price prediction based on deep learning
integration method. Computational intelligence and neuroscience, 2022, 1–9. Https://doi.org/10.1155/2022/1265837.
[15] Samaddar, m., roy, r., de, s., & karmakar, r. (2021, february 19). A comparative study of different machine learning
algorithms on bitcoin value prediction. 2021 international conference on advances in electrical, computing,
communication, and sustainable technologies (icaect). Https://doi.org/10.1109/icaect49130.2021.9392629.
[16] Di tollo, g., andria, j., & filograsso, g. (2023, august 8). The predictive power of social media sentiment: evidence
from cryptocurrencies and stock markets using nlp and stochastic anns. Mathematics, 11(16), 3441.
Https://doi.org/10.3390/math1116344.

452