Showing 22 results for Data
M.a Pohrhoseingholi, H Alavi Majd, A.r Abadi, S Parvanehvar,
Volume 1, Issue 1 (12-2005)
Abstract
Background and Objectives: Missing data exist in many studies, e.g. in regression models, and they decrease the model's efficacy. Many methods have been suggested for handling incomplete data: they have generally focused on missing outcome values. But covariate values can also be missing.
Materials and Methods: In this paper we study the missing imputation by the EM algorithm and auxiliary variable and compare the result with case-complete analysis in a logistic regression model dealing with factors that influence the choice of the delivery method.
Our data came from a cross-sectional study of factors associated with the choice of the delivery method in pregnant women. The sample size in this cross-sectional study was 365 and the data were collected through interviews, using questionnaires covering several demographic variables, delivery history, attitude, and some social factors. We used standard deviations to compare the efficiency of the two methods.
Results: The results show that maximum likelihood analysis by EM algorithm is more effective than case-complete analysis.
The problem of missing data is common in surveys and it causes bias and decreased model efficacy. Here we show that the EM algorithm for imputation in logistic regression with missing values for a discrete covariate is more effective than case-complete analysis.
Conclusion: On the other hand if missing values occur for a continuous covariate then we have to use other methods or change the variable into a discrete one.
N Zare, M Sayadi, E Rezaeyan Fard, H Ghaem,
Volume 6, Issue 1 (6-2010)
Abstract
Background & objectives: statistical modeling explicates the observed changes in data by means of mathematics equations. In cases that dependent variable is count, Poisson model is applied. If Poisson model is not applicable in a specific situation, it is better to apply the generalized Poisson model. So, our emphasis in this study is to notice the data structure, introducing the generalized Poisson regression model and its application in estimates of effective factors coefficients on the number of children and comparing it with Poisson regression model results.
Methods: Besides introducing Poisson regression model, we introduced its application in fertility data analysis. A sample of 1019 women in rural areas of Fars was selected by cross sectional and stratified sampling methods. The number of children of family was determined as a count response variable for model validation.
Results: The sample mean and sample variance of the response variable Y, the number of children, are respectively 4.3 and 8.3 (over-dispersion). Log-likelihood was -1950.93 for Poisson regression and -1946.93 for generalized Poisson regression model.
Conclusions: The results revealed that this data have over-dispersion. According to selection criteria, the suitable model for this data analysis was generalized Poisson regression model. It can estimate effective factors coefficients on the number of children exactly.
Ma Akhoond, A Kazemnejad, E Hajizadeh, Sr Fatemi, A Motlagh,
Volume 6, Issue 4 (3-2011)
Abstract
Background & objectives: Competing risk data is one of the multivarite survival data. Competing risk data can be modelled using copula function. In this study we propose a bayesian modelling approach of competing risk data using the copula function.
Methods: We used the data from colorectal cancer registyrarty in Tehran. After constructing likelihood function using Clayton copula by choosing appropriate prior distribution for parameters, we obtained the posterior distribution of parameters using the Metropolis-Hastings algorithms and Slice sampling.
Results: The results of univariate analysis showed that sex, histology of tumor, extent of wall penetration, lymph node metastasis, distant metastasis and pathological stage of tumor were significantly associated with colon cancer and sex, histology of tumor, lymph node metastasis, distant metastasis and pathological stage of tumor were were significantly related to rectal cancer. In the multivariate analysis, age at diagnosis, tumor grade and distant metastasis were significant prognostic factors for colon cancer and tumor grade and size of the tumor were significant prognostic factors of rectal cancer
Conclusions: As we showed some variables may have different impacts on colon and rectum cancers, consequently, further studies are needed to be conducted considering risk factors of these cancers separately.
Aa Haghdoost, Mr Baneshi, M Marzban,
Volume 7, Issue 2 (9-2011)
Abstract
In the previous paper, the basic concepts of sample size calculation were presented. This paper explores main post-calculation adjustments of the sample size calculation in special circumstances such as multiple group comparisons, unbalanced studies (with unequal number of subjects in different groups) sample size correction for missing data, and adjustment for finite population size. In addition, the concept of design effect in multi-stage sampling
N Mahdavi, M Movahedi, A Khosravi, Y Mehrabi, M Karami, ,
Volume 8, Issue 3 (12-2012)
Abstract
Background and Objectives: Due to the importance of mortality
statistics for planning, setting priorities and equal allocation of health
services in population it is essential to assess quality of reporting
mortality data in health systems. The aim of this study was to evaluate the
completeness and accuracy of the Iranian Vital Horoscope reports for maternal
and the under-five mortality (U5M) in rural areas through its comparison with
other data sources in Iran.
Methods: The mortality data of Vital
Horoscope reported from 30 selected cities over country was compared with the
related data obtained from other data sources including Vital Horoscope's
Fieldwork reports, Death Registration System and Maternal Mortality
Surveillance System of Ministry of Health and Medical Education.
Results: Overall completeness of Vital
Horoscope's Fieldwork reports for U5M in rural areas was about % 62.1. In terms
of cause of death in children under-five,estimated sensitivity values were %
47.2 (95% CI: 22.9-72.2), % 66.6(95% CI: 22.7-95.7), %78.2 (95% CI: 64.3-89.3)for respiratory
infections, diarrhea and vomiting, and injuries-burning and poisoning respectively.
The vital horoscope reports had 12.5% misclassification in determining the cause of maternal death.
Conclusion: Our findings indicate the Vital Horoscope's data might
need some corrections because of underestimating of the mortality indicators.
The comparison of this source with Death Registration System report for causes
of death in children under-five (reported by Vital Horoscope) suggests that the
vital horoscope might have suboptimal quality.
Normal
0
false
false
false
EN-US
X-NONE
AR-SA
N Mahdavi, M Movahedi, A Khosravi, Y Mehrabi, M Karami,
Volume 8, Issue 3 (12-2012)
Abstract
Background and Objectives: Due to the importance of mortality statistics for planning, setting priorities and equal allocation of health services in population it is essential to assess quality of reporting mortality data in health systems. The aim of this study was to evaluate the completeness and accuracy of the Iranian Vital Horoscope reports for maternal and the under-five mortality (U5M) in rural areas through its comparison with other data sources in Iran.
Methods: The mortality data of Vital Horoscope reported from 30 selected cities over country was compared with the related data obtained from other data sources including Vital Horoscope's Fieldwork reports, Death Registration System and Maternal Mortality Surveillance System of Ministry of Health and Medical Education.
Results: Overall completeness of Vital Horoscope's Fieldwork reports for U5M in rural areas was about % 62.1. In terms of cause of death in children under-five,estimated sensitivity values were % 47.2 (95% CI: 22.9-72.2), % 66.6(95% CI: 22.7-95.7), %78.2 (95% CI: 64.3-89.3)for respiratory infections, diarrhea and vomiting, and injuries-burning and poisoning respectively. The vital horoscope reports had 12.5% misclassification in determining the cause of maternal death.
Conclusion: Our findings indicate the Vital Horoscope's data might need some corrections because of underestimating of the mortality indicators. The comparison of this source with Death Registration System report for causes of death in children under-five (reported by Vital Horoscope) suggests that the vital horoscope might have suboptimal quality.
K Holakouie Naieni , S Hashemi Nazari , M Mahmoodi, M Shekari, A Madani,
Volume 9, Issue 4 (3-2014)
Abstract
Background & Objectives: One of the major questions in epidemiological and social science researches is studying the relationship of the living place with social and health outcomes. In this study, we measured segregation indices for a number of important socioeconomic indices using the 2006 Iranian census data to find out whether residential segregation is correlated with the available differences in the health level in the subgroups of certain variables.
Methods: Twenty percent of the 2006 national census data was used for measuring segregation indices. Residential segregation indices were measured once for Hormozgan Province among its eleven cities and then for each city among theirs sections. Six segregation indices were measured. We used the common cut points for interpreting the values of dissimilarity index and information theory index.
Results: According to the dissimilarity and information theory index, the segregation of most of the variables in the province fell within the category of mild segregation. Segregation of the variables in some cities fell within the category of moderate, severe, and even extreme. Conclusion: The results indicated improper distribution of some of these variables in geographic units in some of the cities of Hormozgan Province. This information can help the authorities who are committed to implementing the health equity and social justice.
P Rezanejad Asl , M Hosseini, S Eftekhary, M Mahmoodi , K Nouri,
Volume 10, Issue 3 (12-2014)
Abstract
Background & Objectives : Longitudinal studies are used in many psychiatric researches to evaluate the effectiveness of treatment. The main characteristic of longitudinal studies is repeated measurements of the patients over time. Since observations from the same patient are not independent from each other, especial statistical methods must be used for analyzing the data. Missing data is an indispensable component in longitudinal. In this study, we examined the effect of comprehensive treatment on social-individual performance in patients with the first episode of psychosis.
Methods : The data was from a clinical trial involving patients who were admitted to the clinics of Roozbeh Hospital between 2006_2008. We employed a random effect model for the analysis of longitudinal ordinal responses with non-monotone missingness using the R software version 3.0.2.
Results: The results showed that comprehensive treatment with follow-up at home, age, and family history of the disease had a significant effect on the social-individual performance of the patients. The estimation of the coefficient of age and its standard deviation were 0.05 and 0.03, respectively. The estimation of the coefficient of family history of the disease was -0.82 with a standard deviation of 0.41, and the coefficient of comprehensive treatment with follow-up at home and its standard deviation, were estimated -1.04 and 0.44, respectively.
Conclusion: The model used in this study showed that the comprehensive treatment with follow-up at home was better because individuals under this type of treatment are more likely to have social-individual performance.
Mr Gohari, F Zayeri, Z Moghadami Fard, N Kholdi,
Volume 10, Issue 4 (3-2015)
Abstract
Background and Objectives : Failure to gain weight (FTG) is one of the predominant health issues in children. The aim of this study is application of longitudinal transition model in determining the prognostic factors for failure to gain weight in children under two years.
Methods: In this study, 363 children under 2 years that were visited at the health centers in the east of Tehran were studied. Samples were selected using the two stage clustering method. The study variables were measured repeatedly in 18 consecutive times. Since the data was longitudinal and are dependent, first order transition model was used to determine the risk factors of failure to gain weight. All analyses conducted in R.
Results : The mean (±sd) birth weight was 3057gr(± 838) and 6.9% of the children weighed less than 2500gr at birth. Moreover, 231 children (63.6 %) had no FTW until 2 years of age while 23 ( 6.3 %) had three or more episodes of FTW. Diarrhea (P<0.001), weaning (P<0.001), catching cold (<0.001), and teething (P<.001) were significant risk factors of failure to gain weight. To measure the association between weight loss and the weight in the previous visit, the logarithm of odds ratios was used that was significant (P=0.039).
Conclusion: The association between two consecutive measurements showed that any failure in weight would affect weight gain in the next period of time and the effect of weight deficiency remains for at least one month.
A Afshari Safavi , H Kazemzadeh Gharechobogh , M Rezaei,
Volume 11, Issue 3 (11-2015)
Abstract
Background and Objectives: Missing data is a big challenge in the research. According to the type of the study and of the variables, different ways have been proposed to work with these data. This study compared five popular imputation approaches in addressing missing data in the questionnaires.
Methods: In this study, 500 questionnaires were used for self-medication in diabetic patients. Missing in the observations was artificially generated by random selection of questions and then deleting them. Five imputation ways included: 1) the mean of the questions, 2) the mean of the person, 3) the mode of the person, 4) linear regression, and 5) EM algorithm. For each method, the mean and standard deviation were compared with imputation. The Spearman correlation coefficient, the percentage of incorrectly classified and kappa statistic were also calculated.
Results: A kappa higher than 0.81 represented almost perfect agreement at 10% missingness. The EM algorithm showed the highest level of agreement with the results of actual data with a Kappa of 0.886. With increasing missingness to 30%, the EM algorithm and the mean of the person showed a rather similar agreement with a Kappa of 0.697 and 0.687, respectively.
Conclusion: In this study, the EM algorithm was the most accurate method for handling missing data in all models. The mean of the person method is easy for handling missing data, especially for most non statisticians.
E Ghasemi, M Barooni, R Dehnavieh, M Jafari Sirizi , Mh Mehrolhassani,
Volume 12, Issue 0 (3-2017)
Abstract
Background and Objectives: Health insurance would guarantee people security against disease and health problems. Given the key role of health insurance in achieving the goals of justice and reducing the out-of-pocket payment, this study aimed to evaluate the performance of Iran health insurance using the DEA model in 2014.
Methods: This was a cross sectional study. The study population included all Iran health insurance organizations. DEA input and output criteria were selected by targeted library and documentary review and the data were collected accordingly. The determinants of efficiency were evaluated using liner regression.
Results: The mean technical, management, and scale efficiency of Iran health insurance head offices was 0.593, 0.761, and 0.721, respectively. Considering the findings, the capacity of efficiency promotion at these head offices was approximately 41%. Regarding technical efficiency, 5 head offices had the maximum efficiency (1), 7 head offices had efficiency between 0.5 and 1, and 19 head offices had efficiency less than 0.5. In addition, the variables of population and total number of institutions had a significant impact on efficiency.
Conclusion: Based on defined variables, input oriented AP-DEA model was appropriate. The results showed a great capacity for increasing technical efficiency in the Iranian health insurance organizations which could be increased by benchmarking efficient and reference organizations and also adjusting their input. For this purpose, downsizing and agility of the Iranian health insurance organizations based on the e-government clause are proposed for administrative system reform.
F Esmaili, Mh Mehrolhassani, M Barooni, R Goudarzi ,
Volume 12, Issue 0 (3-2017)
Abstract
Background and Objectives: Productivity and efficiency are the most important and the most common mechanisms of evaluation and measurement of the performance of an enterprise including the Social Security Organization. In the past decades, performance evaluation of various economic sectors has been attractive to researchers in different disciplines. Thus, the aim of this study was to measure the efficiency of the direct treatment section of treatment management units of Social Security Organization by data envelopment analysis method.
Methods: This descriptive- analytic study was conducted to measure the efficiency of the direct treatment section of treatment management units of Social Security Organization through the data envelopment analysis (DEA) method in 2014. Data and relevant statistics were collected from the Statistical Center of Social Security Organization. The Deap2.1 software was used to calculate the efficiency and the EMS software was used to calculate the super efficiency. Then, the hypotheses of the research were studied using the Stata software.
Results: The average technical efficiency, managerial efficiency, and scale efficiency in 2014 was 0.924, 0.992, and 0.932, respectively. Twelve decision making units (DMUs) had the maximum technical efficiency (1), 16 DMUs had technical efficiency between 0.8 and 1, and 3 units had technical efficiency less than 0.8.
Conclusion: This study introduces a functional pattern to managers of Social Security Organization that enables them to have more accurate planning for the development and saving of resources.
R Goudarzi, Mh Mehrolhassani, R Dehnavieh, A Darvishi,
Volume 12, Issue 0 (3-2017)
Abstract
Background and Objectives: Efficiency measurement can be used for all decision-makers and planners for useful resource allocation. Social Security Organization as a health service provider, provides part of health care services in indirect sector. This study aimed to assess the performance of provincial units of Social Security Organization in the indirect health services sector.
Methods: This descriptive analytical study was conducted based on the available data of Social Security Organization in 2014. To assess efficiency and super-efficiency, the DEA-VRS and Anderson-Peterson rating model were used, respectively. Factors affecting the efficiency was evaluated using multivariate regression.
Results: The primary efficacy analysis showed that 61% of the provincial units of Social Security had maximum efficiency. Average efficiency was 0.94. After super-efficiency analysis, it was found that Markazi and Kerman provincial units were the most efficient and the most inefficient units, respectively. Additionally, none of the variables had a significant impact on the efficiency.
Conclusion: Evaluation of the general performance of provincial units reflected the good state of technical efficiency in the indirect health service sector. On the other hand, the scale efficiency of inefficient units compared with managerial efficiency had a higher share of inefficiency. Optimum performance can be achieved through modification of managerial practices for optimal utilization of resources and factors.
K Etemad, A Heidari, Mh Panahi, M Lotfi, F Fallah, S Sadeghi,
Volume 13, Issue 3 (12-2017)
Abstract
Background and Objectives: Data plays a major role in a health care system in development planning and health services support if they are correct, timely and accessible. The data of the Ministry of Health are not readily available and the limited access reduces their value. The aim of this study was to explore the challenges of access to the data of the Iranian Ministry of Health.
Methods: This qualitative study was conducted in 2015. Twenty-three academic and administrative experts were selected purposefully. Semi-structured interviews were conducted to collect the data. The transcripts of the interviews were analyzed using content analysis.
Results: The results of this study provided 4 main themes (challenges of access to the data of the surveillance system, challenges of access to the data of national surveys and ordered projects, challenges of access to the data of electronic health records, and challenges of access to confidential data) and 15 sub-themes.
Conclusion: Given the multiple challenges of access to the data of the Iranian Ministry of Health, it is suggested to design access mechanisms in a systematic manner in the form of guidelines and organizational structures for data access management.
M Amini, A Kazemnejad, F Zayeri , M Gholami Fesharaki,
Volume 13, Issue 4 (3-2018)
Abstract
Background and Objectives: Shift work could threaten health in the long term. The present research aimed to assess the association between shift work and body mass index (BMI) using the multilevel (hierarchical) model during a particular period of time.
Methods: The data of this longitudinal study were collected from a sample of Esfahan’s Mobarakeh steel and Polyacryl companies personnel during 2008 to 2011. Shift work schedule included day work and rotational shift work. The multilevel regression model was utilized for analysing the data and assessing the effect of shift work on BMI by controlling confounding variables including marital status, work expectation, age, company, and educational level.
Results: In this study, of 1368 workers, 42.3% (n=578) and 57.7% (n=790) were day workers and rotating shift workers, respectively. The mean (±SD) age of the day workers and rotating shift workers was 33.07 (±8.66) years and 33.31 (±8.70) years, respectively. After adjusting for confounding variables in a two-level hierarchical model, the association between shift work and BMI was not statistically significant (P=0.837). About 90% of total variation was related to personnel.
Conclusion: According to the results of the present study, no statistically significant relationship was found between shift work schedule and BMI. Thus, other similar studies with a longer follow up period (more than four years) and controlling more confounder factors are necessary to evaluate the relationship between shift work and BMI more accurately.
Mh Mehrolhassani, R Goudarzi, V Yazdi Feyzabadi , Ss Pourhosseini, A Darvishi,
Volume 14, Issue 0 (1-2019)
Abstract
Background and Objectives: Improving the efficiency and productivity of the higher education, especially in the field of research on health sciences, is one of the characteristics of sustainable development in today's societies. This study aimed to measure the efficiency and productivity of Iran's Medical Sciences Universities (MSU) in the research function.
Methods: In a descriptive study, the research function of fourty five MSUs in Iran was evaluated using data envelopment analysis (DEA) method and Malmquist index in 2010, 2013, and 2016 years. Measurement of both efficiency and Malmquist index was developed and modeled based on the assumption of variable returns to scale (VRS) and output-oriented. Also, the ranking of efficient units was done using Anderson-Patterson's model.
Results: The mean research efficiency was estimated to be 0.86. Findings of Malmquist index showed that between 2010 and 2013, there was a 6% growth in the productivity; while the performance of universities had a 12% drop in research function from 2013 to 2016. Also, the average total productivity during two periods is 0.96, indicating 4% reduction in research efficiency which technology efficiency growth has dropped by 8% and other components of total productivity had a positive growth.
Conclusion: The results of the study showed that universities do not work efficiently and average productivity has been decreasing which was mainly due to a decline in the efficiency of technology, which despite the development of technology in recent years could be the result of the lack of effective use of it.
Mh Mehrolhassani, R Goudarzi, V Yazdi Feyzabadi, Ss Pourhosseini, A Darvishi,
Volume 14, Issue 0 (1-2019)
Abstract
Background and Objectives: The higher education system plays an important role in the socio-economic development of the country due to its mission in training the required human resources. Therefore, performance evaluation of different sectors of higher education is of great importance. The present study was conducted to evaluate the educational efficiency and productivity changes of Iranian medical sciences universities.
Methods: This descriptive study was conducted in 2011, 2014, and 2017 to evaluating the performance of 43 Iranian medical universities using Data Envelopment Analysis and output oriented approach. In addition, productivity changes were measured using the Malmquist index. For this purpose, Deap 2.1 software was used. The Anderson Patterson Model and EMS software were also used to rate the units accurately.
Results: The average educational efficiency of medical universities was 0.97 in the study years. The average total productivity based on the Malmquist Index was 1.05, and educational productivity of the universities showed an average growth of 5% over the study years. This growth was 1% from 2011 to 2014 and 10% from 2011 to 2017.
Conclusion: The results of the study showed the acceptable efficiency of the education sector of Iranian medical sciences universities. Moreover, a positive increasing trend was observed in the productivity of the education sector during the study years. Further research using quality and quantity measures are necessary to assess the educational performance of medical universities more accurately.
S Setareh, M Zahiri Esfahani , M Zare Bandamiri , A Raeesi, R Abbasi,
Volume 14, Issue 1 (6-2018)
Abstract
Background and Objectives: Colon cancer is the third most common cancer in the world and the fourth most common cancer in Iran. It is very important to predict the cancer outcome and its basic clinical data. Due to to the high rate of colon cancer and the benefits of data mining to predict survival, the aim of this study was to survey two widely used machine learning algorithms, Bagging and Support Vector Machines (SVM), to predict the outcome of colon cancer patients.
Methods: The population of this study was 567 patients with stage 1-4 of colon cancer in Namazi Radiotherapy Center, Shiraz in 2006-2011. Three hundred and thirty eight patients were alive and 229 patients were dead. We used the Support Vector Machines (SVM) and Bagging methods in order to predict the survival of patients with colon cancer. The Weka software ver 3.6.10 was used for data analysis.
Results: The performance of two algorithms was determined using the confusion matrix. The accuracy, specificity, and sensitivity of the SVM was 84.48%, 81%, and 87%, and the accuracy, specificity, and sensitivity of Bagging was 83.95%, 78%, and 88%, respectively.
Conclusion: The results showed both algorithms have a high performance in survival prediction of patients with colon cancer but the Support Vector Machines has a higher accuracy.
F Ebrahimzadeh, E Hajizadeh, M Birjandi, S Feli, Sh Ghazi,
Volume 14, Issue 3 (12-2018)
Abstract
Background and Objectives: Academic failure is of paramount importance for medical students because it might lead to a decline in scientific level of the community of physicians in the future. This study was conducted to investigate the predictors of academic failure in medical students of Lorestan University of Medical Sciences using classification tree.
Methods: In this cohort study, academic records of all medical students of Lorestan University of Medical Sciences during the academic years of 1999-2008 were selected by census and were followed up until September 2016. Academic failure was defined as having at least one of the components of appropriate grade point average, prolonged graduation, academic probation, dropout, expulsion, and any failure in ccomprehensive exams and the CART classification tree was adopted using the SPSS 22 software to predict it.
Results: The cumulative incidence of academic failure was 26.4% and the most prevalent components were prolonged graduation (21.7%) and academic probation (15.0%). The probability of academic failure was 0.449 in subjects taking guest courses, 0.220 in subjects with no history of guest courses admitted to courses with less than 40 students and admission quotas of zone 1 or 3, and 0.456 in subjects with no history of guest courses admitted to courses with more than 40 students and males.
Conclusion: With respect to identifying the predictors of academic failure, it is suggested that these students be referred to consulting centers of the university or educational supervisors’ moreover, the regulations of taking guest courses in other universities should be revised.
Nasrin Talkhi, Nooshin Akbari Sharak, Zahra Rajabzadeh, Maryam Salari, Seyed Masoud Sadati, Mohammad Taghi Shakeri,
Volume 18, Issue 3 (12-2022)
Abstract
Background and Objectives: Due to the high prevalence of COVID-19 disease and its high mortality rate, it is necessary to identify the symptoms, demographic information and underlying diseases that effectively predict COVID-19 death. Therefore, in this study, we aimed to predict the mortality behavior due to COVID-19 in Khorasan Razavi province.
Methods: This study collected data from 51, 460 patients admitted to the hospitals of Khorasan Razavi province from 25 March 2017 to 12 September 2014. Logistic regression and Neural network methods, including machine learning methods, were used to identify survivors and non-survivors caused by COVID-19.
Results: Decreased consciousness, cough, PO2 level less than 93%, age, cancer, chronic kidney diseases, fever, headache, smoking status, and chronic blood diseases are the most important predictors of death. The accuracy of the artificial neural network model was 89.90% in the test phase. Also, the sensitivity, specificity and area under the rock curve in this model are equal to 76.14%, 91.99% and 77.65%, respectively.
Conclusion: Our findings highlight the importance of some demographic information, underlying diseases, and clinical signs in predicting survivors and non-survivors of COVID-19. Also, the neural network model provided high accuracy in prediction. However, medical research in this field will lead to complementary results by using other methods of machine learning and their high power.
