Stock Price Prediction using artificial intelligence.pptx
1.
Department of ISE
AccurateForecasting for Investment Decisions
Pavani L (1TJ23IS040)
Srushti Shettar (1TJ23IS055)
Uday Kiran B R (1TJ23IS058)
Guide Name : Dr.DoddeGowda
Designation : Associate Professor
Mini Project
STOCK PRICE PREDICTION
2.
01. Introduction
02. LiteratureReview
03. Existing System
04. Problem Statement
05. Proposed System
06. Flow Chart
07. Implementation
08. Results
09. Future Enhancements
10. Conclusion
Agenda
3.
Introduction to StockPrice Prediction
• Definition: Stock price prediction refers to forecasting the future value of company
stocks traded on financial markets.
• Importance: Helps investors, traders, and institutions make informed buy/sell/hold
decisions.
• Data Sources: Historical stock prices, financial statements, market indices, news articles,
and social media sentiment.
• Applications:
• Algorithmic trading
• Portfolio management
4.
Sl.No Title andAuthors Problem Statement Studies and
Contributions
Implementation
Results
Overall
Findings
Future work
01. Prediction in
Stock Price
Using Python
and Machine
Learning
Authors:
Anchal Vij 1 ,
Komal Saxena2
Ajay Rana3
The study
addresses the
challenge of
predicting stock
market trends and
closing prices by
applying machine
learning
algorithms to
historical
financial data. It
seeks to improve
prediction
reliability and aid
investors in
making informed
trading decisions.
The research
reviews prior
stock prediction
techniques and
contributes by
developing and
testing a
Python-based
machine
learning model
that enhances
accuracy and
computational
efficiency. It
demonstrates
the potential of
ML in financial
forecasting
applications.
The
implementation
demonstrated
that machine
learning
algorithms can
effectively
predict stock
prices with a
satisfactory
level of
precision using
historical data.
Machine
learning
proves to be
a valuable
tool in
forecasting
stock prices,
reducing
manual
prediction
errors, and
supporting
better
financial
planning.
Future
research can
focus on
integrating
deep learning
models and
real-time data
for more
dynamic and
accurate stock
market
predictions.
5.
Sl.No Title andAuthors Problem Statement Studies and
Contributions
Implementation
Results
Overall
Findings
Future work
02. Prediction in
Stock Price
Using Machine
Learning
Techniques
Authors :
Sumeet Sarode,
Harsha G.
Tolani, Prateek
Kak, Lifnac
The research
aims to predict
stock closing
prices using
machine learning
techniques
implemented in
Python to assist
investors in
making informed
financial
decisions.
This study
analyzes
previous stock
prediction
approaches and
proposes a
Python-based
machine
learning model
that improves
accuracy and
ease of
implementation.
The
implementation
using historical
datasets shows
that machine
learning
algorithms such as
linear regression
and random forest
yield promising
results in
predicting future
stock prices. The
model produced
consistent and
realistic
predictions
compared to
traditional
methods.
The study
confirms that
machine
learning can
effectively
model complex
stock price
movements and
outperform
basic statistical
approaches. It
also highlights
that data
preprocessing
and algorithm
choice
significantly
influence
prediction
quality.
Future work
could integrate
deep learning
models, real-time
data, and hybrid
predictive
systems for
enhanced market
forecasting
accuracy.
Incorporating
news sentiment
and social media
data is also
recommended to
capture market
psychology.
6.
Sl.No Title andAuthors Problem Statement Studies and
Contributions
Implementation
Results
Overall
Findings
Future work
03. The Directionality
Function Defect of
Performance
Evaluation Method
in Regression
Neural Network
for Stock Price
Prediction
Authors :
Yi Wei
Vipin Chaudhary
The paper addresses
the issue that
existing neural
network evaluation
metrics, such as
prediction error, fail
to capture the
directional
movement (rise or
fall) of stock prices.
It emphasizes that
this lack of
directionality can
mislead financial
decisions and result
in unreliable
performance
evaluations.
The authors
analyze the
limitations of
conventional
performance
metrics like MAE,
MSE, RMSE, and
R² in regression
neural networks.
They demonstrate
that absolute
value–based
evaluation
methods lack
financial
interpretability,
proposing that
directionality
should be
incorporated into
model assessment.
Experiments were
conducted using
20 Chinese and 20
U.S. stocks,
comparing six
neural network
models (MLP,
RNN, LSTM,
GRU, BiRNN,
BiLSTM). Results
show that all
prediction error
values fail to
represent stock
rise or fall trends,
proving that
current evaluation
metrics lack
directional insight.
The study
concludes that
while regression
neural networks
can predict price
closeness
effectively, their
evaluation
methods are
fundamentally
flawed because
they ignore
directional
attributes.
Therefore,
relying solely
on prediction
error values is
unreliable for
financial
forecasting.
Future research
aims to resolve
the defect by
integrating
fractal trend
analysis into
prediction error
evaluation. The
goal is to design
an improved
performance
metric that
accounts for both
price magnitude
and directionality
in stock
prediction
models.
7.
Existing System
01.Statistical Approaches
Statisticalmethods analyze historical stock price data, volume, and related indicators to
identify trends and make predictions
• Time Series Analysis
• Linear Regression
• GARCH (Generalized Autoregressive Conditional Heteroskedasticity)
• Exponential Smoothing
Limitations
• Poor Handling of Non-Stationary Data
• Short-Term Effectiveness Only
• Over fitting Risk
• Volatility Clustering Issues
8.
Existing System
02.Deep Learningin Stock Price Prediction
Deep learning uses multi-layered neural networks to automatically extract complex
patterns and relationships from large volumes of financial data
• Pattern Recognition
• Sequential Modeling
• Integration of Multiple Data Sources
Limitations
• Lack of Interpretability
• Computation & Training Cost
• Regulatory and Ethical Concerns
•Lack of Interpretability
•Lack of Interpretability
•Lack of Interpretability
9.
Problem Statement
• Thestock market is highly dynamic, volatile, and uncertain, with prices
influenced by multiple factors such as historical data, trading volumes,
economic indicators, global events, and market sentiment. Traditional
methods often struggle to capture these complex and non-linear
relationships
• Robust predictive system powered by machine learning in python which
can analyze vast amounts of data, detect hidden patterns, and forecast
stock prices more effectively. The ultimate goal of this system is to help
investors and financial institutions make informed decisions, reduce risks,
and maximize returns.
10.
Proposed System
• DataCollection
• Data Preprocessing
• Prediction & Output
• Website Interface
• Machine Learning Models
Future Enhancement
• CombineARIMA with Deep Learning models (like LSTM or GRU) to capture
both linear and non-linear stock trends.
• Include real-time stock market data for continuous and dynamic prediction
updates.
• Add sentiment analysis from financial news and social media to improve
forecast accuracy.
• Integrate technical and economic indicators (volume, moving averages, GDP,
etc.) for better feature representation.
• Develop a user-friendly web or mobile interface using Flask or Streamlit for
interactive visualization.
• Enable automated model retraining to adapt to changing market patterns.
20.
Conclusion
• The Python-basedstock price prediction model using the ARIMA algorithm
effectively forecasts future stock prices by analyzing historical data and
identifying time-series patterns. ARIMA’s strength lies in its ability to model
linear trends and seasonality, making it a reliable statistical approach for
short-term predictions.
• However, since stock markets are influenced by many non-linear and
unpredictable factors, the model’s accuracy may vary with sudden market
fluctuations. Overall, the project demonstrates how machine learning and
statistical modeling in Python can support investors in making informed
financial decisions and serve as a foundation for more advanced hybrid
forecasting systems in the future.