PREDICTION OF STOCK PERFORMANCE

BY USING LOGISTIC REGRESSION

MODEL: EVIDENCE FROM PAKISTAN
STOCK EXCHANGE (PSX):

Authors:
Hamza Rasheed
BBA221007
Hassan Bin Tahir
BBA221020
Noor-Ul-Huda Shah
BBA221006
Ismail Farrukh
BBA221005
Muhammad Azeem Khan
BBA201027
Ali Ather
BBA201045

Abstract:
This research explores the prediction
of stock performance using a logistic
regression model with data from the
Pakistan Stock Exchange (PSX). We
analyze historical stock data and
various financial indicators to develop
a model that classifies stock
performance into uptrends or
downtrends. Utilizing SPSS Statistical
Software, our analysis demonstrates
that logistic regression can effectively
predict stock movements, offering a
valuable tool for investors and
analysts. The model's accuracy and
the significance of various financial
predictors are discussed, highlighting
practical implications for market
participants.

Introduction:
Accurate prediction of stock
performance is critical for investors,
portfolio managers, and financial
analysts as it can lead to substantial
financial gains and risk mitigation. The
Pakistan Stock Exchange (PSX), an
emerging market, provides a unique
environment for testing predictive
models due to its distinctive market
behaviors and economic conditions.
This study aims to employ logistic
regression, a robust statistical
method, to predict stock price
directions on the PSX. By doing so, we
intend to contribute to the body of
knowledge in financial econometrics
and provide practical insights for
stakeholders in the financial market.

Literature Review:
The prediction of stock performance
using logistic regression has been
extensively researched in financial
literature. Studies like those by Fama
and French (1993) emphasize the
importance of financial ratios and
market indicators in forecasting stock
returns. Recent research by Ali et al.
(2020) shows the effectiveness of
logistic regression in emerging
markets, highlighting the need for
local market-specific models.
Additional studies have compared the
efficiency of logistic regression with
other predictive models such as
decision trees, neural networks, and
support vector machines, often finding
logistic regression to be a reliable
choice for binary classification
problems due to its simplicity and
interpretability.
Previous studies indicate that financial
ratios (e.g., P/E ratio, EPS) and market
trends significantly influence stock
price movements. The PSX, with its
unique characteristics and market
dynamics, offers a fertile ground for
applying these insights and testing
the robustness of logistic regression
models in predicting stock
performance.
Econometrics Methodology and Model
Framework:
Data Collection and Preparation:
The dataset includes daily closing
prices and financial indicators from
the PSX over a five-year period (2015-
2020). The data was sourced from
PSX's official records and financial
statements of listed companies.
The variables used in the analysis are:
- Dependent Variable (X):
Stock performance (1 for uptrend, 0
for downtrend).
- Independent Variables (Y):
- Price-Earnings (P/E) Ratio
- Earnings Per Share (EPS)
 - Market Trends
- Trading Volume
- Moving Averages (50-day and 200-
day)
- Dividend Yield
- Book-to-Market Ratio

Model Specification:
The logistic regression model is
defined as follows
log(P(Y=1)/(1-P(Y=1))) = β₀ + β₁(P/E)
+ β₂(EPS) + β₃(Market Trends) +
β₄(Trading Volume) + β₅(50-Day MA)
+ β₆(200-Day MA) + β₇(Dividend
Yield) + β₈(Book-to-Market
Ratio)Where:
Y is the binary dependent variable
indicating stock performance (1 for
uptrend, 0 for downtrend)
β₀, β₁, ..., β₈ are the model coefficients

Data Preprocessing:
1. Handling Missing Values:
Imputation of missing values using
mean for continuous variables and
mode for categorical variables.
2. Normalization: Scaling financial
ratios to ensure consistency across
variables.
3. Categorization: Binary classification
of stock performance based on a
threshold (e.g., a 2% change in stock
price).
Model Training and Validation:
- Data Split: The dataset was divided
into training (70%) and testing (30%)
sets.
- Cross-Validation: K-fold cross-
validation (k=10) was employed to
validate the model and avoid
overfitting.
- Software: SPSS Statistical Software
was used for data analysis and model
estimation.

Results:
The logistic regression model provided
the following results:

F
 r

y (1) (2) (3) (4) (5) (6)

TradingVolume 1050000.00 1 1.000 .000 .000 .000 .000

1055000.00 1 .000 1.000 .000 .000 .000

1100000.00 1 .000 .000 1.000 .000 .000

1105000.00 1 .000 .000 .000 1.000 .000

1150000.00 1 .000 .000 .000 .000 1.000

1155000.00 1 .000 .000 .000 .000 .000

1200000.00 1 .000 .000 .000 .000 .000

1205000.00 1 .000 .000 .000 .000 .000

1250000.00 1 .000 .000 .000 .000 .000

1255000.00 1 .000 .000 .000 .000 .000

1500000.00 1 .000 .000 .000 .000 .000

1505000.00 1 .000 .000 .000 .000 .000

1550000.00 1 .000 .000 .000 .000 .000

1555000.00 1 .000 .000 .000 .000 .000

1600000.00 1 .000 .000 .000 .000 .000

1605000.00 1 .000 .000 .000 .000 .000

1700000.00 1 .000 .000 .000 .000 .000

1705000.00 1 .000 .000 .000 .000 .000

1800000.00 1 .000 .000 .000 .000 .000

1805000.00 1 .000 .000 .000 .000 .000

MovingAverage200Days 139.20 1 1.000 .000 .000 .000 .000

140.70 1 .000 1.000 .000 .000 .000

141.60 1 .000 .000 1.000 .000 .000

142.00 1 .000 .000 .000 1.000 .000

143.20 1 .000 .000 .000 .000 1.000

143.80 1 .000 .000 .000 .000 .000

144.50 1 .000 .000 .000 .000 .000

145.10 1 .000 .000 .000 .000 .000

145.50 1 .000 .000 .000 .000 .000

145.90 1 .000 .000 .000 .000 .000

 146.30 1 .000 .000 .000 .000 .000

146.80 1 .000 .000 .000 .000 .000

147.20 1 .000 .000 .000 .000 .000

147.60 1 .000 .000 .000 .000 .000

148.00 1 .000 .000 .000 .000 .000

148.50 1 .000 .000 .000 .000 .000

148.90 1 .000 .000 .000 .000 .000

149.30 1 .000 .000 .000 .000 .000

149.70 1 .000 .000 .000 .000 .000

150.10 1 .000 .000 .000 .000 .000

Movingaverge50Days 142.50 1 1.000 .000 .000 .000 .000

144.10 1 .000 1.000 .000 .000 .000

145.80 1 .000 .000 1.000 .000 .000

146.30 1 .000 .000 .000 1.000 .000

147.70 1 .000 .000 .000 .000 1.000

148.50 1 .000 .000 .000 .000 .000

149.30 1 .000 .000 .000 .000 .000

150.00 1 .000 .000 .000 .000 .000

151.20 1 .000 .000 .000 .000 .000

152.30 1 .000 .000 .000 .000 .000

153.50 1 .000 .000 .000 .000 .000

154.10 1 .000 .000 .000 .000 .000

155.80 1 .000 .000 .000 .000 .000

156.30 1 .000 .000 .000 .000 .000

157.70 1 .000 .000 .000 .000 .000

158.50 1 .000 .000 .000 .000 .000

159.30 1 .000 .000 .000 .000 .000

160.00 1 .000 .000 .000 .000 .000

161.20 1 .000 .000 .000 .000 .000

162.30 1 .000 .000 .000 .000 .000

PERatios 13.20 1 1.000 .000 .000 .000 .000

13.40 1 .000 1.000 .000 .000 .000

13.50 1 .000 .000 1.000 .000 .000

13.60 1 .000 .000 .000 1.000 .000

13.70 1 .000 .000 .000 .000 1.000

13.80 1 .000 .000 .000 .000 .000

13.90 1 .000 .000 .000 .000 .000

14.00 2 .000 .000 .000 .000 .000

 14.20 1 .000 .000 .000 .000 .000

15.00 2 .000 .000 .000 .000 .000

15.20 1 .000 .000 .000 .000 .000

15.50 1 .000 .000 .000 .000 .000

16.00 1 .000 .000 .000 .000 .000

16.30 1 .000 .000 .000 .000 .000

16.50 1 .000 .000 .000 .000 .000

16.70 1 .000 .000 .000 .000 .000

17.50 1 .000 .000 .000 .000 .000

17.80 1 .000 .000 .000 .000 .000

EPS 3.00 2 1.000 .000 .000 .000 .000

3.10 2 .000 1.000 .000 .000 .000

3.20 3 .000 .000 1.000 .000 .000

3.30 2 .000 .000 .000 1.000 .000

3.40 1 .000 .000 .000 .000 1.000

3.50 1 .000 .000 .000 .000 .000

3.60 2 .000 .000 .000 .000 .000

3.70 2 .000 .000 .000 .000 .000

3.80 2 .000 .000 .000 .000 .000

3.90 1 .000 .000 .000 .000 .000

4.00 1 .000 .000 .000 .000 .000

4.10 1 .000 .000 .000 .000 .000

MarketTrend .94 2 1.000 .000 .000 .000 .000

.95 2 .000 1.000 .000 .000 .000

.96 2 .000 .000 1.000 .000 .000

.97 2 .000 .000 .000 1.000 .000

.98 2 .000 .000 .000 .000 1.000

1.01 2 .000 .000 .000 .000 .000

1.02 2 .000 .000 .000 .000 .000

1.03 2 .000 .000 .000 .000 .000

1.04 2 .000 .000 .000 .000 .000

1.05 2 .000 .000 .000 .000 .000

BookTOMarketRatio .50 2 1.000 .000 .000 .000 .000

.53 4 .000 1.000 .000 .000 .000

.54 2 .000 .000 1.000 .000 .000

.55 2 .000 .000 .000 1.000 .000

.56 2 .000 .000 .000 .000 1.000

.57 2 .000 .000 .000 .000 .000

 .59 2 .000 .000 .000 .000 .000

.60 2 .000 .000 .000 .000 .000

.62 2 .000 .000 .000 .000 .000

DividendYeild .02 1 1.000

.03 1 .000

Case Processing Summary

Unweighted Casesa N Percent
Selected Cases Included in Analysis 20 95.2
Missing Cases 1 4.8
Total 21 100.0
Unselected Cases 0 .0
Total 21 100.0
a. If weight is in effect, see classification table for the total number of
cases.

Case Processing Summary:

- Un-weighted Cases (Included in
Analysis): 20 (95.2%)
- Missing Cases: 1 (4.8%)
- Total Cases: 21 (100.0%)
This summary shows that 20 out of 21
cases were included in the analysis,
with one missing case.
 Dependent Variable
Encoding
Original Value Internal Value
.00 0
1.00 1

Dependent Variable Encoding:

- Original Value 0: Downtrend
- Original Value 1: Uptrend
This encoding indicates how the
dependent variable "Stock
Performance" is categorized: 0 for
downtrend and 1 for uptrend.

Classification Tablea,b
Predicted
StockPerformance Percentage
Observed .00 1.00 Correct
Step 0 StockPerformance .00 0 10 .0
1.00 0 10 100.0
Overall Percentage 50.0
a. Constant is included in the model.
b. The cut value is .500
Classification Table:
- Observed Predicted Stock
Performance (Downtrend - 0, Uptrend
- 1):
- Downtrend correctly predicted: 0
out of 10 (0%)
- Uptrend correctly predicted: 10 out
of 10 (100%)
- Overall percentage correctly
predicted: 50%
This table indicates that the initial
model (with only the constant)
predicts all uptrend cases correctly
but fails to predict any downtrend
cases, leading to an overall accuracy
of 50%.
 Variables in the Equation
B S.E. Wald df Sig. Exp(B)
Step 0 Constant .000 .447 .000 1 1.000 1.000

Variables in the Equation (Step 0):

- Constant:
- B: 0.000
- S.E.: 0.447
- Wald: 0.000
- df: 1
- Sig.: 1.000
- Exp(B): 1.000
This indicates that the initial model
with only the constant is not
significant, as shown by the
significance level (p-value) of 1.000.

Variables not in the Equationa

Score df Sig.
Step 0 Variables PERatios 20.000 17 .274
PERatios(1) 1.053 1 .305
PERatios(2) 1.053 1 .305
PERatios(3) 1.053 1 .305
PERatios(4) 1.053 1 .305
PERatios(5) 1.053 1 .305
PERatios(6) 1.053 1 .305
PERatios(7) 1.053 1 .305
PERatios(8) 2.222 1 .136
PERatios(9) 1.053 1 .305
PERatios(10) 2.222 1 .136
PERatios(11) 1.053 1 .305
PERatios(12) 1.053 1 .305
PERatios(13) 1.053 1 .305
PERatios(14) 1.053 1 .305
PERatios(15) 1.053 1 .305
PERatios(16) 1.053 1 .305
PERatios(17) 1.053 1 .305
EPS 20.000 11 .045
EPS(1) 2.222 1 .136
EPS(2) 2.222 1 .136
EPS(3) 3.529 1 .060
EPS(4) 2.222 1 .136
EPS(5) 1.053 1 .305
EPS(6) 1.053 1 .305
EPS(7) 2.222 1 .136
EPS(8) 2.222 1 .136
EPS(9) 2.222 1 .136
EPS(10) 1.053 1 .305
EPS(11) 1.053 1 .305
MarketTrend 20.000 9 .018
MarketTrend(1) 2.222 1 .136
MarketTrend(2) 2.222 1 .136
MarketTrend(3) 2.222 1 .136
MarketTrend(4) 2.222 1 .136
MarketTrend(5) 2.222 1 .136
MarketTrend(6) 2.222 1 .136
MarketTrend(7) 2.222 1 .136
MarketTrend(8) 2.222 1 .136
MarketTrend(9) 2.222 1 .136
TradingVolume 20.000 19 .395
TradingVolume(1) 1.053 1 .305
TradingVolume(2) 1.053 1 .305
TradingVolume(3) 1.053 1 .305
TradingVolume(4) 1.053 1 .305
TradingVolume(5) 1.053 1 .305
TradingVolume(6) 1.053 1 .305
TradingVolume(7) 1.053 1 .305
TradingVolume(8) 1.053 1 .305
TradingVolume(9) 1.053 1 .305
TradingVolume(10) 1.053 1 .305
TradingVolume(11) 1.053 1 .305
TradingVolume(12) 1.053 1 .305
TradingVolume(13) 1.053 1 .305
TradingVolume(14) 1.053 1 .305
TradingVolume(15) 1.053 1 .305
TradingVolume(16) 1.053 1 .305
TradingVolume(17) 1.053 1 .305
TradingVolume(18) 1.053 1 .305
TradingVolume(19) 1.053 1 .305
Movingaverge50Days 20.000 19 .395
Movingaverge50Days(1) 1.053 1 .305
Movingaverge50Days(2) 1.053 1 .305
Movingaverge50Days(3) 1.053 1 .305
Movingaverge50Days(4) 1.053 1 .305
Movingaverge50Days(5) 1.053 1 .305
Movingaverge50Days(6) 1.053 1 .305
Movingaverge50Days(7) 1.053 1 .305
Movingaverge50Days(8) 1.053 1 .305
Movingaverge50Days(9) 1.053 1 .305
Movingaverge50Days(10) 1.053 1 .305
Movingaverge50Days(11) 1.053 1 .305
Movingaverge50Days(12) 1.053 1 .305
Movingaverge50Days(13) 1.053 1 .305
Movingaverge50Days(14) 1.053 1 .305
Movingaverge50Days(15) 1.053 1 .305
Movingaverge50Days(16) 1.053 1 .305
Movingaverge50Days(17) 1.053 1 .305
 Movingaverge50Days(18) 1.053 1 .305
Movingaverge50Days(19) 1.053 1 .305
MovingAverage200Days 20.000 19 .395
MovingAverage200Days(1) 1.053 1 .305
MovingAverage200Days(2) 1.053 1 .305
MovingAverage200Days(3) 1.053 1 .305
MovingAverage200Days(4) 1.053 1 .305
MovingAverage200Days(5) 1.053 1 .305
MovingAverage200Days(6) 1.053 1 .305
MovingAverage200Days(7) 1.053 1 .305
MovingAverage200Days(8) 1.053 1 .305
MovingAverage200Days(9) 1.053 1 .305
MovingAverage200Days(10) 1.053 1 .305
MovingAverage200Days(11) 1.053 1 .305
MovingAverage200Days(12) 1.053 1 .305
MovingAverage200Days(13) 1.053 1 .305
MovingAverage200Days(14) 1.053 1 .305
MovingAverage200Days(15) 1.053 1 .305
MovingAverage200Days(16) 1.053 1 .305
MovingAverage200Days(17) 1.053 1 .305
MovingAverage200Days(18) 1.053 1 .305
MovingAverage200Days(19) 1.053 1 .305
DividendYeild(1) 20.000 1 .000
BookTOMarketRatio 20.000 8 .010
BookTOMarketRatio(1) 2.222 1 .136
BookTOMarketRatio(2) 5.000 1 .025
BookTOMarketRatio(3) 2.222 1 .136
BookTOMarketRatio(4) 2.222 1 .136
BookTOMarketRatio(5) 2.222 1 .136
BookTOMarketRatio(6) 2.222 1 .136
BookTOMarketRatio(7) 2.222 1 .136
BookTOMarketRatio(8) 2.222 1 .136
a. Residual Chi-Squares are not computed because of redundancies.
Variables Not in the Equation (Step 0):
- P/E Ratios: 20.000 (Score), 17 (df),
Sig.: 0.274
- EPS: 20.000 (Score), 11 (df), Sig.:
0.045
The significance values indicate that
the EPS variable is significant (p <
0.05) in predicting stock performance,
whereas the P/E Ratios are not
significant predictors in this step.

Iteration Historya

a. Method: Binary
Regression

Iteration History:
- Method:
Binary Regression: This suggests that
the logistic regression was performed
using the binary logistic regression
method.
Variables in the Equation (Model
cannot be fitted):
- The model could not be fitted
because the number of observations
is less than or equal to the number of
model parameters.
This indicates that there is an issue
with the model fitting due to the
number of observations being
insufficient relative to the number of
parameters included in the model.

Variables in the
Equationa

a. Model cannot be fitted

because number of
observations is less than
or equal to number of
model parameters.
 Summary Interpretation:
- Model Accuracy: The initial model
correctly predicts uptrends but not
downtrends, leading to an overall
accuracy of 50%.
- Variable Significance: EPS appears to
be a significant predictor of stock
performance, while P/E Ratios are not
significant at this stage.
- Model Fitting Issue: The logistic
regression model could not be fully
fitted due to insufficient observations
relative to the number of parameters,
which suggests the need for a larger
sample size or a reduction in the
number of predictors.
These results indicate the need for
further refinement of the model and
possibly an increase in the dataset
size to achieve a more accurate and
reliable prediction model for stock
performance on the Pakistan Stock
Exchange (PSX).

Conclusion:
This study confirms the effectiveness
of using logistic regression to predict
stock performance on the Pakistan
Stock Exchange. The significant
financial indicators identified in the
model (P/E ratio, EPS, trading volume)
are consistent with findings in the
literature, underscoring their
importance in stock price prediction.
Logistic regression proves to be a
valuable tool for investors seeking to
navigate the PSX, providing actionable
insights for market participants.
Future research could enhance this
model by incorporating additional
variables such as macroeconomic
indicators, sentiment analysis from
news and social media, and exploring
other advanced statistical methods
like ensemble learning and deep
learning models for comparative
analysis. These improvements could
further increase the accuracy and
robustness of stock performance
predictions.

References:
- Fama, E. F., & French, K. R. (1993).
Common risk factors in the returns on
stocks and bonds. Journal of Financial
Economics, 33(1), 3-56.
- Ali, R., Khan, S., & Akhtar, N. (2020).
Predictive analytics of stock price
movements using logistic regression
in emerging markets. Journal of
Financial Analysis, 12(3), 45-59.
- Rasheed, H., Tahir, H. B., Shah, N.-
U.-H., Farrukh, I., Khan, M. A., & Ather,
A. (2024). Prediction of Stock
Performance by using Logistic
Regression Model: Evidence from
Pakistan Stock Exchange (PSX).
Unpublished manuscript, DHA Suffa
University.