© 20XX IJNRD | Volume X, Issue X Month 20XX | ISSN: 2456-4184 | IJNRD.
ORG
Prediction of Cognitive Events from Physiological
Data of Pilots Using Machine Learning Algorithms
1
Swapnil M Wanjare, 2Dr. V.T. Gaikwad
1
Research Scholar, 2Professor
1
Electronics,
1
Sipna College of Engineering and Technology, Amravati, India
Abstract : The aviation sector is always looking for ways to increase security and lower deaths. Flight-related mishaps are caused by a lack
of situational awareness while in flight, and airplane accidents are typically disastrous. Proposed a model to identify the cognitive states of pilots
from their psychophysiological signals to warn them before slipping into a risky mental state, such as channelled attention, diverted attention,
and startle/surprise. Utilizing machine learning algorithms, which can analyze enormous volumes of data to find trends and anticipate future
safety hazards, is one promising strategy. By examining data from numerous sources, including flight logs, maintenance logs, and
meteorological data, this research investigates the potential of machine learning techniques to lower aviation fatalities. The goal of the study is
to create a machine learning algorithm that can identify potential dangers and notify pilots and ground personnel accordingly. The study
explores these algorithms' usage in aviation safety from an ethical standpoint and suggests some rules for doing so. This study ultimately shows
how machine learning algorithms can help to lower airline fatalities.Utilize such measurements to differentiate between their cognitive states
using classification techniques like logistic regression, random forest, and support vector machines. The outcomes could be used as a component
of the risk management system to notify pilots when needed. These models will be developed in Python and majorly focus on Ensemble
Learning Algorithms along with classifications.
IndexTerms - Machine Learning, Artificial Intelligence, Ensemble Methods, Random Forest, Bagging, Boosting.
________________________________________________________________________________________________________
I. INTRODUCTION
INTRODUCTION
Hospital or health care waste is generally named & popular as biomedical waste. The world health organization defines
biomedical waste as ,’’Waste generation by health care activities & includes blood, used needles, pharmaceuticals, radioactive
materials etc.” The biomedical waste is also known as infectious waste or medical waste or health care waste. According to
biomedical waste management & handling rules 1998 of India. Biomedical waste means any waste which is generated during the
diagnosis, treatment or immunization of human being or animals or in research activities. In simple words biomedical waste is the
waste generated by the medical & health institute/agencies.
Biomedical waste management defines waste management as the practices & procedures or the administration of activities that
provide for the collection, source separation, storage, transportation, transfer, processing, treatment & disposal of waste .
Biomedical waste management is a routine procedure of hospital administration as prescribed by law .Hospital waste , hospital
acquired infection , transfusion transmitted diseases, rising incidence of hepatitis B, HIV & Other diseases, create potential threat
of infection, contamination & serious health hazards to doctors, nurses, ward boys, support staff, sanitation workers, rag pickers
& other health care workers. Who are regularly exposed to biomedical waste as an occupation hazards as well as general public in
the surrounding area .
NEED OF THE STUDY.
The establishment of large hospitals where hundreds to thousands of patients are treated , it has created a serious problems of
biomedical waste management. The seriousness of improper biomedical waste management was brought to the light during
summer 1998. In India studies have been carried out at local / regional levels in various hospitals, indicate that roughly about 1-5
kg/bed/day to waste is generated. Among all health care personnel ,ward boys , sweepers, operation theatre & laboratory
attendants have come into contact with biomedical waste during the process of segregation , collection, transport, storage & final
disposal . The knowledge of medical , paramedical staff & ward boys , sweepers about the biomedical waste management is
important to improve the biomedical waste management practices. The biomedical waste requiring special attention includes
PaperID 25
International Journal of Novel Research and Development (www.ijnrd.org)
© 20XX IJNRD | Volume X, Issue X Month 20XX | ISSN: 2456-4184 | IJNRD.ORG
those that are potentially infectious , sharps ,example needle , scalpels , objects capable of puncturing the skin , also
plastic ,pharmaceutical & chemically hazardous substances used in laboratories etc.
3.1Population and Sample
KSE-100 index is an index of 100 companies selected from 580 companies on the basis of sector leading and market
capitalization. It represents almost 80% weight of the total market capitalization of KSE. It reflects different sector company’s
performance and productivity. It is the performance indicator or benchmark of all listed companies of KSE. So it can be regarded
as universe of the study.Non-financial firms listed at KSE-100 Index (74 companies according to the page of KSE visited on
20.5.2015) are treated as universe of the study and the study have selected sample from these companies.
The study comprised of non-financial companies listed at KSE-100 Index and 30 actively traded companies are selected
on the bases of market capitalization.And 2015 is taken as base year for KSE-100 index.
3.2 Data and Sources of Data
For this study secondary data has been collected. From the website of KSE the monthly stock prices for the sample firms
are obtained from Jan 2010 to Dec 2014. And from the website of SBP the data for the macroeconomic variables are collected for
the period of five years. The time series monthly data is collected on stock prices for sample firmsand relative macroeconomic
variables for the period of 5 years. The data collection period is ranging from January 2010 to Dec 2014. Monthly prices of KSE -
100 Index is taken from yahoo finance.
3.3 Theoretical framework
Variables of the study contains dependent and independent variable. The study used pre-specified method for the
selection ofvariables. The study used the Stock returns are as dependent variable. From the share price of the firm the Stock
returns are calculated. Rate of a stock salable at stock market is known as stock price.
RESEARCH METHODOLOGY
The methodology section outline the plan and method that how the study is conducted. This includes Universe of the study,
sample of the study,Data and Sources of Data, study’s variables and analytical framework. The detailsare as follows;
3.1Population and Sample
KSE-100 index is an index of 100 companies selected from 580 companies on the basis of sector leading and market
capitalization. It represents almost 80% weight of the total market capitalization of KSE. It reflects different sector company’s
performance and productivity. It is the performance indicator or benchmark of all listed companies of KSE. So it can be regarded
as universe of the study.Non-financial firms listed at KSE-100 Index (74 companies according to the page of KSE visited on
20.5.2015) are treated as universe of the study and the study have selected sample from these companies.
The study comprised of non-financial companies listed at KSE-100 Index and 30 actively traded companies are selected
on the bases of market capitalization.And 2015 is taken as base year for KSE-100 index.
3.2 Data and Sources of Data
For this study secondary data has been collected. From the website of KSE the monthly stock prices for the sample firms
are obtained from Jan 2010 to Dec 2014. And from the website of SBP the data for the macroeconomic variables are collected for
the period of five years. The time series monthly data is collected on stock prices for sample firmsand relative macroeconomic
variables for the period of 5 years. The data collection period is ranging from January 2010 to Dec 2014. Monthly prices of KSE -
100 Index is taken from yahoo finance.
3.3 Theoretical framework
Variables of the study contains dependent and independent variable. The study used pre-specified method for the
selection ofvariables. The study used the Stock returns are as dependent variable. From the share price of the firm the Stock
returns are calculated. Rate of a stock salable at stock market is known as stock price.
Systematic risk is the only independent variable for the CAPM and inflation, interest rate, oil prices and exchange rate
are the independent variables for APT model.
Consumer Price Index (CPI) is used as a proxy in this study for inflation rate. CPI is a wide basic measure to
computeusualvariation in prices of goods and services throughout a particular time period. It is assumed that arise in inflation is
inversely associated to security prices because Inflation is at lastturned into nominal interest rate andchange in nominal interest
rates caused change in discount rate so discount rate increase due to increase in inflation rate and increase in discount rateleads
todecreasethe cash flow’s present value (Jecheche, 2010). The purchasing power of money decreased due to inflation, and due to
which the investors demand high rate of return, and the prices decreased with increase in required rate of return (Iqbal et al,
2010).
Exchange rate is a rate at which one currency exchanged with another currency. Nominal effective exchange rate (Pak
Rupee/U.S.D) is taken in this study.This is assumed that decrease in the home currency is inverselyassociated to share prices
(Jecheche,2010). Pan et al. (2007) studied exchange rate and its dynamic relationship with share prices in seven East Asian
Countries and concludethat relationshipof exchange rate and share prices varies across economies of different countries. So there
PaperID 26
International Journal of Novel Research and Development (www.ijnrd.org)
© 20XX IJNRD | Volume X, Issue X Month 20XX | ISSN: 2456-4184 | IJNRD.ORG
may be both possibility of either exchange rate directly or inverselyrelated with stock prices.Oil prices are positively related with
share prices if oil prices increase stock prices also increase (Iqbal et al, 1012).Ataullah (2001) suggested that oil prices cause
positive change in the movement of stock prices. The oil price has no significant effect on stock prices (Dash & Rishika,
2011).Six month T-bills rate is used as proxy of interest rate. As investors arevery sensitive about profit and where the signals
turn into red they definitely sell the shares. And this sensitivity of the investors towards profit effects the relationship of the stock
prices and interest rate, so the more volatility will be there in the market if the behaviors of the investors are more sensitive.
Plethora (2002)has tested interest rate sensitivity to stock market returns, and concluded an inverse relationship between interest
rate and stock returns. Nguyen (2010) studies Thailand market and found thatInterest rate has aninverse relationship with stock
prices.
KSE-100 index is used as proxy of market risk. KSE-100 index contains top 100 firms which are selected on the bases of
their market capitalization. Beta is the measure of systematic risk and has alinear relationship with return (Horn, 1993). High risk
is associated with high return (Basu, 1977, Reiganum, 1981 and Gibbons, 1982). Fama and MacBeth (1973) suggested the
existence of a significant linear positive relation between realized return and systematic risk as measured by β. But on the other
side some empirical results showed that high risk is not associated with high return (Michailidis et al. 2006, Hanif, 2009). Mollah
and Jamil (2003) suggested thatrisk-return relationship is notlinear perhaps due to high volatility.
3.4Statistical tools and econometric models
This section elaborates the proper statistical/econometric/financial models which are being used to forward the study
from data towards inferences. The detail of methodology is given as follows.
3.4.1 Descriptive Statistics
Descriptive Statics has been used to find the maximum, minimum, standard deviation, mean and normally distribution of
the data of all the variables of the study. Normal distribution of data shows the sensitivity of the variables towards the periodic
changes and speculation. When the data is not normally distributed it means that the data is sensitive towards periodic changes
and speculations which create the chances of arbitrage and the investors have the chance to earn above the normal profit. But the
assumption of the APT is that there should not be arbitrage in the market and the investors can earn only normal profit. Jarque
bera test is used to test the normality of data.
3.4.2 Fama-Mcbeth two pass regression
After the test statistics the methodology is following the next step in order to test the asset pricing models. When testing
asset pricing models related to risk premium on asset to their betas, the primary question of interest is whether the beta risk of
particular factor is priced. Fama and McBeth(1973)develop a two pass methodology in which the beta of each asset with respect
to a factor is estimated in a first pass time series regression and estimated betas are then used in second pass cross sectional
regression to estimate the risk premium of the factor. According to Blum (1968) testing two-parameter models immediately
presents an unavoidable errors-in-the variables problem.It is important to note that portfolios (rather than individual assets) are
used for the reason of making the analysis statistically feasible.Fama McBeth regression is used to attenuate the problem of
errors-in-variables (EIV) for two parameter models (Campbell, Lo and MacKinlay, 1997).If the errors are in the β (beta)of
individual security are not perfectly positively correlated, the β of portfolios can be much more precise estimates of the true β
(Blum, 1968).
The study follow Fama and McBeth two pass regressionto test these asset pricing models.The Durbin Watson is used to
check serial correlation and measures the linear association between adjacent residuals from a regression model. If there is no
serial correlation, the DW statistic will be around 2. The DW statistic will fall if there is positive serial correlation (in worst case,
it will be near zero). If there is a negative correlation, thestatistic will lie somewhere between 2 and 4. Usually the limit for non-
serial correlation is considered to be DW is from 1.8 to 2.2. A very strong positive serial correlation is considered at DW lower
than 1.5 (Richardson and smith, 1993).
According to Richardson and smith(1993) to make the model more effective and efficient the selection criteria for the
shares in the period are: Shares with no missing values in the period, Shares with adjusted R 2 < 0 or F significant (p-value)
>0.05of the first pass regression of the excess returns on the market risk premium are excluded. And Shares are grouped by
alphabetic order into group of 30 individual securities (Roll and Ross, 1980).
3.4.2.1 Model for CAPM
In first pass the linear regression is used to estimate beta which is the systematic risk.
Ri−R f =( Rm−R f ) β (3.1)
Where RiisMonthly return of thesecurity, Rf isMonthly risk free rate, Rm isMonthly return of market and βis systematic risk
(market risk).
The excess returns Ri - Rf of each security is estimated from a time series share prices of KSE-100 index listed shares for
each period under consideration. And for the same periodthe market Premium R m - Rfalso estimated. After that regress the excess
returns Ri - Rf on the market premium Rm - Rfto find the beta coefficient (systematic risk).
Then a cross sectional regression or second pass regression is used on average excess returns of the shares and estimated betas.
Ȓi=γ 0+ γ 1 β 1+ є(3.2)
Where ƛ0= intercept, ȒIis average excess returns of security i,βIisestimated be coefficient of security I and Є is error term.
PaperID 27
International Journal of Novel Research and Development (www.ijnrd.org)
© 20XX IJNRD | Volume X, Issue X Month 20XX | ISSN: 2456-4184 | IJNRD.ORG
3.4.2.2 Model for APT
In first pass the betas coefficients are computed by using regression.
Ri−R f =βi f 1+ β i 2 f 2 + β i 3 f 3 + β i 4 f 4 + ϵ (3.3)
Where Ri is the monthly return of stock i,Rf is risk free rate, βi is the sensitivity of stock i with factors and ϵ is the error term.
Then a cross sectional regression or second pass regression is used on average excess returns of the shares on the factor scores.
Ȓ=γ 0+ γ 1 β 1+ γ 2 β 2+ γ 3 β 3 +γ 4 β 4 +ϵ i (3.4)
WhereȒ is average monthly excess return of stock I, ƛ = risk premium, β1 to β4 are the factors scores and εi is the error term.
3.4.3 Comparison of the Models
The next step of the study is to compare these competing models to evaluate that which one of these models is more
supported by data.This study follows the methods used by Chen (1983), the Davidson and Mackinnon equation (1981) and the
posterior odds ratio (Zellner, 1979) for comparison of these Models.
3.4.3.1 Davidson and MacKinnon Equation
CAPM is considered the particular or strictly case of APT. These two models are non-nested because by imposing a set
of linear restrictions on the parameters the APT cannot be reduced to CAPM. In other words the models do not have any common
variable. Davidson and MacKinnon (1981) suggested the method to compare non-nested models. The study used the Davidson
and MacKinnon equation (1981) to compare CAPM and APT.
This equation is as follows;
Ri=α R APT + ( 1−α ) RCAPM +e i (3.5)
WhereRi= the average monthly excess returns of the stock i, R APT= expected excess returns estimated by APT, R CAPM= expected
excess returns estimated by CAPM and α measure the effectiveness of the models. The APT is the accurate model to forecast the
returns of the stocks as compare to CAPMif α is close to 1.
3.4.3.2 Posterior Odds Ratio
A standard assumption in theoretical and empirical research in finance is that relevant variables (e.g stock returns) have
multivariate normal distributions (Richardson and smith, 1993). Given the assumptionthat the residuals of the cross-sectional
regression of the CAPM and the APT satisfy the IID (Independently and identically distribution) multivariate normal assumption
(Campbell, Lo and MacKinlay, 1997), it is possible to calculate the posterior odds ratio between the two models.In general the
posterior odds ratio is a more formal technique as compare to DM equation and has sounder theoretical grounds (Aggelidis and
Maditinos, 2006).
The second comparison is done using posterior odd radio. The formula for posterior odds is given by Zellner (1979) in favor of
model 0 over model 1.
The formula has the following form;
N /2
R=[ ESS0 / ESS1 ] N K −K /2 (3.6)
0 1
WhereESS0iserror sum of squares of APT, ESS1iserror sum of squares of CAPM, Nisnumber of observations, K 0is
number of independent variables of the APT and K 1 isnumber of independent variables of the CAPM.As according to the ratio
when;
R> 1 means CAPM is more strongly supported by data under consideration than APT.
R < 1 means APT is more strongly supported by data under consideration than CAPM.
IV. RESULTS AND DISCUSSION
4.1 Results of Descriptive Statics of Study Variables
Std. Jarque-Bera test Sig
Variable Minimum Maximum Mean Deviation
KSE-100 Index
-0.11 0.020 0.047
0.14 5.558 0.062
Inflation -0.01 0.02 0.007 0.008 1.345 0.510
Exchange rate -0.07 0.04 0.003 0.013 1.517 0.467
Oil Prices -0.24 0.11 0.041 0.060 2.474 0.290
Interest rate -0.13 0.05 0.047 0.029 1.745 0.418
Table 4.1: Descriptive Statics
Table 4.1 displayed mean, standard deviation, maximum minimum and jarque-bera test and its p value of the macroeconomic
variables of the study. The descriptive statistics indicated that the mean values of variables (index, INF, EX, OilP and INT) were
0.020, 0.007, 0.003, 0.041 and 0.047 respectively. The maximum values of the variables between the study periods were 0.14,
0.02, 0.04, 0.41, 0.11 and 0.05 for the KSE- 100 Index, inflation, exchange rate, oil prices and interest rate.
The standard deviations for each variable indicated that data were widely spread around their respective means.
PaperID 28
International Journal of Novel Research and Development (www.ijnrd.org)
© 20XX IJNRD | Volume X, Issue X Month 20XX | ISSN: 2456-4184 | IJNRD.ORG
Column 6 in table 4.1 shows jarque bera test which is used to checkthe normality of data. The hypotheses of the normal
distribution are given;
H0 : The data is normally distributed.
H1 :The data is not normally distributed.
Table 4.1 shows that at 5 % level of confidence, the null hypothesis of normality cannot be rejected. KSE-100 index and
macroeconomic variables inflation, exchange rate, oil prices and interest rate are normally distributed.
The descriptive statistics from Table 4.1 showed that the values were normally distributed about their mean and variance. This
indicated that aggregate stock prices on the KSE and the macroeconomic factors, inflation rate, oil prices, exchange rate, and
interest rate are all not too much sensitive to periodic changes and speculation. To interpret, this study found that an individual
investor could not earn higher rate of profit from the KSE. Additionally, individual investors and corporations could not earn
higher profits and interest rates from the economy and foreign companies could not earn considerably higher returns in terms of
exchange rate. The investor could only earn a normal profit from KSE.
Table 1 Table Type Styles
Table
TableColumnHead
Head
Tablecolumnsubhead Subhead Subhead
copy Moretablecopya
II. ACKNOWLEDGMENT
Thepreferredspellingoftheword “acknowledgment” inAmericaiswithoutan “e” afterthe “g”.Avoidthestiltedexpression,
“Oneofus(R.B.G.)thanks...”
Instead,try“R.B.G.thanks”.Putapplicablesponsoracknowledgmentshere;DONOTplacethemonthefirstpageofyourpaperorasafootnote.
REFERENCES
[1] Ali, A. 2001.Macroeconomic variables as common pervasive risk factors and the empirical content of the Arbitrage Pricing
Theory. Journal of Empirical finance, 5(3): 221–240.
[2] Basu, S. 1997. The Investment Performance of Common Stocks in Relation to their Price to Earnings Ratio: A Test of the
Efficient Markets Hypothesis. Journal of Finance, 33(3): 663-682.
[3] Bhatti, U. and Hanif. M. 2010. Validity of Capital Assets Pricing Model.Evidence from KSE-Pakistan.European Journal of
Economics, Finance and Administrative Science, 3 (20).
PaperID 29
International Journal of Novel Research and Development (www.ijnrd.org)
