Proceedings of the 2nd International Conference on Financial Technology and Business Analysis

DOI: 10.54254/2754-1169/45/20230263

Using Machine Learning Methods to Predict Tesla Stock

Silong Dai1,a,*
1
Chengdu University of Technology, Sichuan Province 610059, China
a. David<[Link]@[Link]>
*corresponding author

Abstract: Electric vehicles are now a common mode of transportation due to the aggressive
promotion of new energy. Tesla currently has the largest market share among all brands, as
well as its techniques and patents may ensure the benefits of future development. And also,
Elon Musk is a powerful and ambitious businessman who can steer a technological company
toward a brighter future. Yet, due to some uncontrollable factors like Covid-19, disaster
incidents, technical staff resignations, etc., might influence the prospects of Tesla stock. So,
in this study, machine learning techniques will be primarily employed to forecast Tesla stock
prices' trajectory over the next 30 days. The two primary methods used to forecast and
evaluate accuracy to determine which model is more appropriate are linear regression and
random forest. Before the model is trained, all of the stock data is divided into a training side
and a test side. According to the research, linear regression model performs better in
predicting the direction of Tesla stock than a random forest model. Based on the search results,
it can be said that machine learning methods are likely to unearth patterns and insights that
humans haven’t seen before and can be used to make accurate and unmistakable stock
predictions.

Keywords: Machine Learning Methods, Tesla Stock, Linear Regression, Random Forest

1. Introduction
Tesla is an American business that primarily creates electric vehicles and solutions for, renewable
energy. In 2003, the battery, computer software, and patentable motor were among the foundational
technologies used by it [1-2]. In an initial public offering, it did so for the first time by an American
carmaker, and in 2010, it won multiple awards in the industry. In nine years, it had already achieved
global expansion and opened Shanghai, China as the site of its first "Gigafactory" outside of the
United States. The apparent worldwide move towards renewable energy is the reason why Tesla’s
stock is parabolic [3–4]. So, it can be said that the electric vehicle manufacturer is in a position to
experience tremendous expansion in the years to come [5]. Tesla is currently the most successful
manufacturer of new energy vehicles. Elon Musk, the tesla company owner, has developed into a
prosperous businessman who can guide Tesla toward a better future. Artificial intelligence, batteries,
and alternative energy are some of their industries. The research demonstrates that these industries
have a very positive tendency for the future [6]. Tesla’s profitability and share price have increased
since going public, in part because of more advanced technology and business practices. The
company has also attracted many investment institutions to purchase shares because electric vehicles
are following the future development trend and its technology and patents are leading in the market.
© 2023 The Authors. This is an open access article distributed under the terms of the Creative Commons Attribution License 4.0
([Link]

95
 Proceedings of the 2nd International Conference on Financial Technology and Business Analysis
DOI: 10.54254/2754-1169/45/20230263

Its share price began at slightly more than $1 and reached a high of more than $400. Therefore, more
and more experts have made predictions about Tesla's stock in light of the swift growth of its price
across the entire financial market. Such fundamental analysis method: Use financial models and
valuation models to forecast Tesla's future development trends and stock prices using the company's
financial data, market conditions, industry development trends, and other factors. News analysis
technique: Predict Tesla's future performance and stock price movements based on an analysis of
news reports and market research studies on the company's rivals. But in terms of these two basic
methods, the company cannot make an accurate prediction for the trend of tesla stock. While it exists
some limits. The aim of the study utilizing machine learning techniques is to more precisely forecast
the growth trend of Tesla stock to make some people get a return on investment.
So, this paper will primarily employ machine learning methods to forecast Tesla stock in the next
30 days. The principle of machine learning techniques is historical market data and corporate financial
data, using machine learning algorithms and artificial intelligence technology to establish a
forecasting model to predict Tesla's stock price. And linear regression algorithm and random forest
algorithm will be mainly used in the research. The last five years' worth of data was chosen by the
data provider, [Link]. The paper begins by scanning the data to see if any values are missing.
The data set's 20% leftover data is used as a test set to verify the model, and then the paper randomly
selects 80% of it to generate a training set. It takes careful computational planning to create
predictions using previous data. Machine Learning can validate the accuracy of our predictions
utilizing the training sample and test sample. And the paper’s primary attention is on the closing price.
The closing price is employed as the research object in the machine learning method for prediction.
The most appropriate parameters are chosen by performing 30 days spectrum analysis with the daily
closing price as the reference volume. The purpose of the study using machine learning ways is to
more precisely predict the growth trend of Tesla stock by comparing two models’ accuracy: linear
regression and random forest. Using RAE, RMSE, MSER, and R2 indicators to judge which model is
better. Based on the analysis and comparison of these two models, the study found that the linear
regression model performs better in predicting the growth trend of Tesla stock. The results suggest
that machine learning algorithms can effectively predict the stock market and provide valuable
insights for investors.
The remainder of this essay is organized as follows. Section 2 of this article will display the data
information, its sources, and how it was transformed into the measures used for analysis. Section 3
of the study presents and describes the findings, emphasizing their significance in the context of war
and economic interdependence. Finally, the paper summarizes the research findings and contributions.
2. Data and Methodology
2.1. Dataset Collection and Processing
Data set collection and processing is a very basic module that manages the selection of appropriate
datasets for the most part. The dataset that will be used to determine market expectations needs to be
sorted in light of different factors. Information preprocessing, a component of information mining, is
transforming unstructured data into something more intelligent. Preprocessing the data entails
checking for missing characteristics, looking for all possible values, dividing the dataset into
manageable portions, and testing the data in preparation for further processing. This project uses the
Tesla dataset from Kaggle, which has 1258 instances and 7 characteristics. In the following Table 1
shows some details data.

96
 Proceedings of the 2nd International Conference on Financial Technology and Business Analysis
DOI: 10.54254/2754-1169/45/20230263

Table 1: Seven characteristics.

Date Open High Low Close Adj Close Volume
1 2016/8/16 45.098 45.438 44.6819 44.722 44.722 11335500
2 2016/8/17 44.8660 44.966 44.5600 44.6479 44.6479 8935500
3 2016/8/18 44.764 45.132 44.458 44.702 44.702 8572500
4 2016/8/19 44.708 45.034 44.5060 45 45 8297500
5 2016/8/22 44.834 45.0219 44.5359 44.5859 44.5859 10327500
6 2016/8/23 44.8639 45.6980 44.5600 44.9679 44.9679 23922000
7 2016/8/24 45.41 45.43 44.444 44.5239 44.5239 12853500

The models are adjusted and fitted using the preparation sets. It is divided into an 80:20 ratio, with
80% going toward the preparation phase and the remaining 20% going toward the testing phase.
Besides, this paper uses used statistical methods to describe the mean, variance, maximum, minimum,
and median of the data set in Table 2.

Table 2: Descriptive statistical analysis.

Open High Low Close Adj Close Volume
mean 183.2740 187.2652 178.9412 183.3511 183.3511 4.441050e+07
min 36.2200 36.9459 35.3979 35.7939 35.7939 8.297500e+06
50% 66.6020 67.9500 65.4459 66.7560 66.7560 3.498150e+07
max 891.3800 900.4000 871.5999 883.0900 883.0900 3.046940e+08

Then this paper respectively uses a linear regression model and random forest model to forecast
the 30-day trend of tesla stock.
2.2. Linear Regression
The relationship between independent a Then this paper respectively uses a linear regression model
and random forest model to forecast the 30-day trend of tesla stock.
And dependent variables are established by employing linear regression, which is a fundamental
machine learning algorithm. By minimizing the gap between the anticipated and actual values to
generate the most effective linear model to predict the future stock value, it focuses its methodology
on the linear relationship among a collection of input factors (throughout factors) and target variables
(dependent variables). It is often used in the field of forecasting and modeling, such as stock price
forecasting, house price forecasting, sales forecasting, traffic forecasting, etc. In addition, linear
regression is also commonly used in feature selection, anomaly detection, data exploration, and other
fields. In addition, business, science, and pretty much every other sector where expectations and
anticipating are important use linear regression. The formula for the linear regression model is as
follows:

Y = b0 + b1X1 + b2X2 +…+bnXn (1)

The predicted y - value, being variable to be predicted, and the independent variables make up the
linear regression model. X1, X2...Xn, representing factors that influence the dependent variable. The
intercept, denoted by b0, and the coefficients of the independent variables, represented by b1, b2...bn,
complete the model. Following figure 1 will show the comparison between actual stock and predicted
stock using a linear regression model. Figure 1 illustrates the number of days, and Y demonstrates
the closing price of the Tesla stock.

97
 Proceedings of the 2nd International Conference on Financial Technology and Business Analysis
DOI: 10.54254/2754-1169/45/20230263

Figure 1: Comparison between actual stock and predict stock in next 30 days.

2.3. Random Forest

In classification problems, a machine learning method known as a random forest is utilized, which
involves constructing multiple decision trees during the training phase. The modeling process of the
random forests is the trees' crucial factor that determines. whereas for regression tasks, the median or
standard deviation forecast of each tree is returned. The propensity of decision trees to extrapolate
their training set can be corrected by random decision forests. [7]. Even though the decision tree
algorithm generally supersedes random forests in accuracy, horizontal stripe trees are more accurate.
However, data characteristics can impact how well they perform. [8-9]. Random forests outperform
bagged trees because of a minor modification that decorrelates the trees. Using boot-strapped training
samples, the paper constructs some decision trees. A random sample of m forecasters is selected as
split possibilities when thinking about just a breakup in a tree from the whole set of p predictors.
Every instance separation is considered, the process is repeated. A new sample of m predictors is
chosen for each split in which only one of the m predictors can be used. It is typically chosen that m
p is used, implying that the calculation of the overall prediction number is roughly corresponding to
the number of predictors taken into account at each split [10]. Random forests are commonly used
for classification and regression problems and perform well in many practical applications such as
financial forecasting, medical diagnosis, image recognition, etc. The formula for the random forest
model is as follows:
1𝑀
𝑦̂ = 𝑀 𝛴ⅈ=1 𝑓ⅈ (𝑥) (2)

Where 𝑦̂ is the outcome of the hypothesis of the input sample x, 𝑓ⅈ (𝑥) shows the prediction result
of the i-th decision tree, and M is the random forest's number of decision trees. The random forest
model will be used in Figure 2 to forecast the trend of Tesla stock over the next 30 days.

98
 Proceedings of the 2nd International Conference on Financial Technology and Business Analysis
DOI: 10.54254/2754-1169/45/20230263

Figure 2: Predict the 30-day trend of Tesla stock using a random forest model.

3. Results
To forecast the trend of tesla's stock over the subsequent 30 days based on the current data set, two
separate models, linear regression and random forest, were used. Although the parameters for the two
models' training and test sets are identical, there is a slight variation in the prediction accuracy of the
two models. While the random forest model's accuracy is only 97.25 percent, that of the linear
regression model is up to 99.99% when compared to the R2. This means that linear regression-based
machine learning techniques are preferable to random forest techniques for data set prediction. In
addition, there are certain differences in the accuracy of MAE, MSE, and MRSE between the two,
and table 3 below displays the precise differences.

Table 3: Prediction accuracy of linear regression model and random forest model.
R2 MAE MSE MRSE

Linear 99.96 1.763 17.986 4.241

regression
Random Forest 97.25 18.972 1301.363 36.074

The findings in the following table demonstrate that both models are excellent fits for the given
data set, but the linear regression model is still superior. First of all, the prediction accuracy of R2 of
the line regression model is close to 1. Secondly, through the above Figure1 and Figure2, it can be
seen intuitively that the regression model using the line between the predicted value and the real value
is more reductive It can be claimed that these two machine learning approaches are used to predict
Tesla stock is desirable given the current research landscape since prediction accuracy reveals that
Tesla's movement in the next 30 days will likely be similar to the outcomes indicated by the two
models. Apart from the R2 value indicating the high predictive accuracy of the first model, it also has
the advantage of interpretability. This indicates that the relationship between the input variables and
the anticipated result in the linear regression approach is simpler to comprehend. This can be useful
in situations where it is important to understand how different factors are affecting the outcome, such
as in financial analysis. Furthermore, the linear regression model has been widely used in various
fields, such as economics, social sciences, and engineering. Therefore, it has a strong theoretical and

99
 Proceedings of the 2nd International Conference on Financial Technology and Business Analysis
DOI: 10.54254/2754-1169/45/20230263

practical foundation, and it is easier to compare the results of the linear regression model with
previous studies or established models. However, it is worth noting that the choice of the machine
learning approach depends on the specific problem and data set being analyzed. When assessing
findings, other methodologies for machine learning much like the random forest or support vector
machines might become preferable. Therefore, it is important to carefully evaluate and compare
different machine learning models before selecting the most suitable one for a given task.
4. Conclusion
According to this study, using machine learning techniques is theoretically feasible to predict the
prices of Tesla stock. The following reasons may help to explain this. Both the random forest and
linear regression models used in this study preprocessed the data before splitting it into two sets. The
final result was mathematically calculated to determine the corresponding R2, MAE, MSE, and
MRSE and use the matplotlib standard library to visualize the data. This research does, however, have
some restrictions. Just linear regression and random forest are used in the machine learning approach.
In addition, it's essential to bear in mind that there are still some restrictions on using algorithms to
guess stock prices. One limitation is the availability and quality of data. Inaccurate or incomplete data
can lead to inaccurate predictions, and some factors that may affect stock prices may not be captured
in the available data. Another drawback is the potential for overfitting, which arises when a model is
excessively intricate and closely fits the training data. The accuracy rate upon that training instances
may arise from this, but performance on fresh, unused data may suffer. To avoid overfitting, carefully
select and preprocess the data, select an appropriate model, and evaluate the model's performance
using techniques such as cross-validation. Consequently, it is only possible to conclude that in the
context of this study, machine learning is very effective at predicting the trend of the stock. Even so,
there is a good chance that machine learning techniques will be used to forecast stock market trends.
These techniques can be applied to the related variable operations mentioned above in the future for
further optimization, or they can be used to compare various models to make it easier to conduct in-
depth research on this subject. To conclude, the potential of machine learning techniques in
forecasting stock prices is evident, however, their efficacy relies on the nature of the problem and
dataset under investigation. Further research is needed to explore the potential of these techniques
and to develop more accurate and reliable models for predicting stock prices.
References
[1] Yahoo!: Tesla, Inc. (TSLA) Stock Price, News, Quote & History. Retrieved April 3, 2023, from
[Link]
[2] Srivastava P., Mishra P, K.: Stock Market Prediction Using RNN [Link] 2021 2nd Global Conference for
Advancement in Technology (GCAT), pp. 1-5. IEEE, Bangalore, India (2021).
[3] Marchai F. L., Martin W., Suhartono D.: Stock Prices Prediction Using Machine Learning. In: 8th International
Conference on Information Technology, Computer and Electrical Engineering (ICITACEE), pp. 79-84. IEEE,
Semarang, Indonesia (2021).
[4] Mottaghi N., Farhangdoost S.: Stock Price Forecasting in Presence of Covid-19 Pandemic and Evaluating
Performances of Machine Learning Models for Time-Series Forecasting, Papers 2105.02785, [Link]. (2021).
[5] Chatterjee A., Bhowmik H., Sen J.: Stock Price Prediction Using Time Series, Econometric, Machine Learning, and
Deep Learning Models. In: 2021 IEEE Mysore Sub Section International Conference (MysuruCon), pp. 289-
[Link] (2021).
[6] Bhuriya D., Kaushal G., Sharma A., Singh U.: Stock Market Prediction Using A Linear Regression. In: International
conference of Electronics, Communication and Aerospace Technology (ICECA), pp. [Link] (2017).
[7] Hastie, T., Tibshirani, R., Friedman, Jerome, F.: The Elements of Statis-tical Learning Data Mining, Inference, and
Prediction. 2nd edn. Springer, Heidelberg (2009).
[8] Piryonesi S, M., El-Diraby T, E.: Role of Data Analytics in Infrastructure Asset Management: Overcoming Data
Size and Quality Problems. Journal of Transportation Engineering, Part B: Pavements. 146 (2), (2020).

100
 Proceedings of the 2nd International Conference on Financial Technology and Business Analysis
DOI: 10.54254/2754-1169/45/20230263

[9] Piryonesi S., M., El-Diraby T, E.: Using Machine Learning to Examine Impact of Type of Performance Indicator
on Flexible Pavement Deterioration Modeling. Journal of Infrastructure Systems. 27 (2), (2021).
[10] Gareth J., Daniela W., Trevor H., et al.: An Introduction to Statistical Learning with applications in R. P343-345,
Springer New York, NY (2021).

101