platforms like Netflix (Singh, 2020), the pandemic
unexpectedly acted as an accelerator. First, the
pandemic restricted travel and offline entertainment
activities, leading a significant number of consumers
to turn to online streaming services to meet their
entertainment needs. Netflix, with its extensive
content library and convenient viewing experience,
successfully attracted a large number of new users
and increased the stickiness of existing users.
Secondly, During the pandemic, the competitive
dynamics of the international streaming market
experienced significant changes. Netflix, by utilizing
its established brand reputation and strategic market
advantages, further reinforced its leading position,
outperforming rivals in viewership and subscriber
growth.
As of now, Exploration on Netflix's stock has
largely focused on normal periods. For example,
Singh and Kumar applied various machine learning
techniques to predict Netflix's stock prices and
evaluated the effectiveness of different techniques.
Alternatively, research has focused on improving
prediction accuracy. For instance, research proposed
a hybrid forecasting method combining ARIMA
models with neural networks to increase the
reliability of predictions for Netflix's stock price by
incorporating more precise data and advanced
analytical techniques (Garcia, 2021).
Considering the gap, this paper plans to focus on
the impact of the coronavirus pandemic as an external
factor on Netflix's stock market prices and explore the
development of the streaming industry in the post-
pandemic era through stock price predictions. By
comparing the accuracy of different machine learning
algorithms, aims to identify the most effective model
for abnormal stock fluctuations. The best model will
be used to forecast Netflix's stock prices, providing
valuable insights for investors and analyzing the
development of the media industry in the post-
pandemic era through the stock trends of Netflix, a
representative streaming enterprise.
2 METHOD
2.1 Preparation
The Netflix stock price dataset on Kaggle used in this
study provides historical data related to Netflix's
stock prices (Kaggle, 2024), often used for financial
analysis, time series forecasting, and data
visualization. This dataset comprises 6,750 data
points over six years, from December 2, 2019, to May
24, 2024. Typically, the data is available in CSV
format, easily imported into data analysis tools.
This study uses a dataset without missing values,
eliminating the need for imputation of missing
closing prices. Outliers, which may indicate
extremely high or low prices, are detected and
addressed using statistical methods such as Z-score or
IQR. To stabilize data variance and mitigate the
impact of extreme values, a moving window
approach smooths the stock price data, replacing the
original closing prices with the 30-day moving
average. The dataset is split into an 80% training set
and a 20% test set, ensuring adequate data for model
training and sufficient data for testing model
generalization. To realistically simulate the model's
performance in practical scenarios, the dataset is
divided into two segments: the training set, which
consists of the initial 80% of the data for model
training and development, and the test set, which
includes the remaining 20% for assessing the model's
accuracy and generalization.
Throughout the observation period, Netflix's
stock price demonstrated significant fluctuations and
growth. The average opening price was $432.54, with
a standard deviation of $528.40, reflecting
considerable uncertainty in opening prices. Similarly,
the highest price, lowest price, and closing price
exhibited comparable volatility, indicating a dynamic
valuation process by the stock market. During the
early pandemic, the global economy was severely
impacted, posing unprecedented challenges for most
industries. However, for streaming platforms like
Netflix, the pandemic acted as an unexpected
accelerator. Charts reveal a significant upward trend
in Netflix's stock price from 2020, with ongoing
fluctuations through 2024. Both opening and closing
prices rose substantially during this period, signalling
market optimism regarding Netflix's future growth
potential.
2.2 Machine Learning-Based Models
This study employs three models—Random Forest,
XGBoost, and LSTM—for comparative analysis,
utilizing historical stock price data to train and
validate the models' predictive performance. The
input data for the task includes historical stock prices,
the output of the task is the predicted stock price value
for a specific future period . To evaluate the
performance of the proposed models, this study used
Mean Absolute Error (MAE), Mean Squared Error
(MSE), and R2 metrics.
2.2.1 LSTM
The architecture of an LSTM includes four essential
components: the input gate, which controls the
integration of new information into the cell state; the
Statistics and Analysis of Netflix Stock Price in the Post-Pandemic Era Based on Machine Learning Algorithms