mechanical engineers

Upload: nestor-lopez

Post on 07-Mar-2016

22 views

Category:

Documents


0 download

DESCRIPTION

Mecánica

TRANSCRIPT

  • PANORAMA OF THE MARKET DEMAND FOR MECHANICAL ENGINEERS IN SOUTH AMERICAN COUNTRIES

    Ronald A. Cardenas

    Faculty of Mechanical Engineering, Universidad Nacional de Ingeniera

    Lima, Peru [email protected]

    Kevin S. Bello Faculty of Mechanical Engineering, Universidad Nacional de Ingeniera

    Lima, Peru [email protected]

    Alexander R. Valle

    Faculty of Mechanical Engineering, Universidad Nacional de Ingeniera

    Lima, Peru [email protected]

    Elizabeth R. Villota Section of Mechanical Engineering, Pontificia Universidad Catlica del

    Peru Lima, Peru

    [email protected]

    Alberto M. Coronado Faculty of Mechanical Engineering, Universidad Nacional de Ingeniera

    Lima, Peru [email protected]

    ABSTRACT

    Many educational institutions employ surveys in order to

    identify what majors to offer or what competencies to

    emphasize in their curricula. Different from a survey, we

    present an analysis of the labor market needs based on data

    collected from job ads available in the Internet. Tools of

    natural language processing (NLP) and statistical techniques

    have been employed to handle the job ads. For Peru, Chile and

    Colombia, a detailed panorama of the market demand has

    been depicted: mechanical engineering appears among the top

    most demanded engineering majors and maintenance is its

    most frequent technical requirement; management (project,

    quality and operations) related requirements also rank high,

    together with a working knowledge of English. By using

    diverse visualization techniques we can also show the social network of a major, where friendship is defined by the amount of job ads shared by any two majors.

    1 INTRODUCTION During the past decade, many South American countries

    have benefited from soaring commodity prices. Along with

    their economic growth, their required working forces have

    also grown in complexity and specialization. Many local

    employers are having a hard time finding qualified workers,

    especially in technical fields [1]. Educational institutions are

    also having difficulty to supply the demanded labor.

    Particularly, for these educational institutions, the challenge of

    providing talented human labor to their local economies is

    now bigger than ever, as commodity prices are down and

    rising productivity is imperative, if sustained growth is desired

    [2].

    For educational institutions around the world, and South

    America is not the exception, it is not always clear what

    majors they should offer or what competencies they should

    emphasize in their curricula; and it is also not self-evident

    what the relative importance should be among foundational,

    professional and emerging (entrepreneurial, innovation) skills

    [3,4,5,6]. Some institutions have adopted the ABET program

    outcomes and guide their curricula design towards meeting

    those outcomes; however, to what extent can a developing

    country use those ABET recommendations? and what about

    incorporating their own economies needs into their

    curriculum? or how much effort do they have to devote to

    innovation? Different countries have different needs,

    resources, cultures, and hence require different solutions [7].

    Proceedings of the ASME 2015 International Mechanical Engineering Congress and Exposition IMECE2015

    November 13-19, 2015, Houston, Texas

    IMECE2015-51557

    1 Copyright 2015 by ASME

  • Considering the specific case of mechanical engineering,

    Wolfe [8] performed a survey among mechanical engineering

    alumni of MIT. It was found that topics closely related to the

    core undergraduate curriculum in mechanical engineering

    were identified as the least used in their daily jobs. On the

    other hand, topics such as, communications, teamwork,

    independent thinking, professional skills and attitudes, and

    personal skills and attitudes, were identified among the most

    commonly used. Interestingly, the topics just mentioned also

    were reported as being learnt mostly outside MIT. In addition,

    three important aspects were stressed: 1) MIT graduates

    pursued a wide variety of careers, 2) mechanical engineers

    needed to learn more than just knowledge about the physical

    world, and 3) engineers needed to have strong business skills,

    such as communication.

    An essential body of knowledge (BOK) for mechanical

    engineering students was proposed by Jarosz & Busch-

    Vishniac [9]. The authors dissected current curricula in

    mechanical engineering, at nine US institutions, into a list of

    required topics. Among the 64 topics deemed essential, 45%

    correspond to engineering, 18% to mathematics, 15% to

    physics, 12% to chemistry, 4% to computer science, 2% to

    communication, 2% to statistics, 1% to economics, 1% to

    ethics, and 0% to biology. The BOK obtained represents

    what is currently taught, not necessarily what should be taught. In addition, it was observed that many of these topics reflect mechanical engineering in the seventeenth and

    eighteenth centuries, not necessarily what is currently needed.

    At present, universities across the world are engaged in a

    race to increase their position in the academic rankings

    currently available. This practice is now under increasing

    scrutiny. Hill et al. [10] discuss the idea that the structure and

    makeup of chemical engineering faculties in the US, and how

    hiring and promotion is carried out, may affect the training of

    undergraduate students regarding industrial applications. In

    the same vein, Xue [11] analyses if there is in fact a STEM crisis or a STEM surplus in the US. It was reported that depending on the major, and sector, it can be one or another.

    For example, in the academic job market, there was no

    noticeable shortage of any discipline. On the other hand, in the

    private sector, there was high demand for software developers,

    petroleum engineers, data scientists and skilled trades.

    Educational institutions tend to appeal to surveys to

    employers, as well as alumni, in order to obtain a glimpse of

    the labor market. Different from a survey, whose results are

    tightly related to the quality and scope of questions performed,

    in this study, we present an analysis of the labor market needs

    based on data collected from job ads (tens of thousands)

    available in the internet. Working with this data, we are able to

    depict a detailed panorama of the market demand in three

    South American countries: Peru, Chile and Colombia. The job

    ads are handled using tools of natural language processing

    (NLP), information retrieval and diverse statistical and

    visualization techniques. Uni-, bi- and tri-gram-based

    language models were used to extract the most common job

    requirements, which can be readily examined as Wordclouds.

    The remainder of this article is organized as follows. The

    next section presents a brief discussion on the techniques used

    to obtain, process and analyze natural language data.

    Following, we present the results of our analysis. We start by

    showing the market demand by country for diverse

    engineering majors. Then we discuss the relationship among

    mechanical engineering and other engineering majors taking

    into account how often these majors share job ads. Next, we

    present Wordclouds with the most common requirements for

    mechanical engineers. The final section presents the

    conclusions.

    2 NATURAL LANGUAGE PROCESSING

    2.1 Reference Corpus The corpus used in this work was built by extracting job

    ads from several popular job search websites from Peru, Chile

    and Colombia. Since the same job ad can be published in more

    than one website, we consider it as repeated if the description

    of the position is found within the last fifteen days in the

    database. The time periods in which the data were extracted

    are mentioned in Section 3.

    This corpus consists of more than 250000 job ads for

    Peru, 180000 for Colombia and 50000 for Chile. It contains

    job ads from all areas in the market as published in the

    websites.

    2.2. Pre-processing Phase There is no standard format for job ads in Latin America.

    The writing style is not formal and the text can surely contain

    orthographic errors. Therefore, the data must be cleaned

    before being classified. Techniques from natural language

    processing (NLP) such as regular expressions (regex) and

    tokenization were used to eliminate special symbols, HTML

    tags, multiple spaces and so on. Furthermore, low frequency

    words were mapped to a small set of tags, such as Number,

    Date, LongWord, Url, PhoneNumber, among others.

    2.3. Majors Classification The classifier was based on regular expressions. It

    searches for major names surrounded by a pre-established set

    of phrases, such as ...professional with major in... or ...student with major in.... The expression to match is constructed joining up to four sets of commonly seen phrases,

    before and after the possible major name, in order to obtain as

    much as possible relationships.

    The 30 most popular engineering majors were accounted

    for graduates from the three countries considered here. The

    classifier relies on multi-word identifiers for each major

    considered, in order to perform multi-classification. Each

    major identifier set was tuned for each country.

    The evaluation of the classifier was conducted in a semi-

    automatic manner, since no annotated data was available. The

    2 Copyright 2015 by ASME

  • evaluation corpus consisted of 1000 randomly sampled job

    ads. The accuracy for the top five most demanded engineering

    majors in Peru can be appreciated in Table 1.

    TABLE 1. Classifier accuracy for the top five most

    demanded engineering majors in Peru

    Engineering major Accuracy (%)

    Industrial 94.01

    Systems 89.49

    Informatics 89.19

    Electrical 88.02

    Mechanical 88.10

    2.4. Market Demand Visualization Two ways of visualizing the market demand were

    selected. The first one, known as Treemap, shows the demand

    per major in a space-constrained manner. Each engineering

    major is represented as a rectangle with area proportional to its

    demand.

    The second one, known as Circleplot, allows us to

    visualize interconnectedness among engineering majors taking

    into account how often these majors share job ads. Each pair

    of interconnected majors is linked by an arch. The arch

    thickness is proportional to the interconnectedness between

    majors, expressed as the number of shared job ads.

    These graphics are shown in Section 3 for each country.

    Interactive versions are available online for Peru [12,13],

    Chile [14,15] and Colombia [16,17].

    2.5. Requirements Ranking and Visualization Job requirements were modeled as keywords and ranked

    using statistical scores. Potential keywords consisted of uni-

    grams, bi-grams and tri-grams, filtered by a list of more than

    3000 corpus tuned stopword phrases. This stopword set was

    built combining standard Spanish stopwords, manually tuned

    stopwords and smart stopwords. The smart stopwords were extracted with statistical techniques mentioned by Rose et al

    [18].

    Three ranking scores were compared. The first one was

    term frequency tft, frequency of the term t in the entire corpus.

    The main source of noise in this score was the fact that a

    phrase can be mentioned multiple times, but only once as a

    requirement. In order to deal with the ambiguity of uni-gram

    terms, we perform a correction in uni-gram counts before

    calculating their scores -only for terms in common with the

    tri-gram list. This correction consisted in adding up the counts

    of the tri-gram phrases in which the uni-gram term being

    analyzed has the same meaning. The semantic relationship

    was manually determined.

    The second one was document frequency dfw, the number

    of ads in the corpus that contains the word w. For this score,

    there was no need for correction in uni-gram counts because

    this score removes the noise from phrases mentioned in

    different contexts, as was the case with the first scoring

    method.

    The third score compared was a discretized version of TF-

    IDF score as implemented by Gutwin and Nevil-Manning

    [19]. For the case of the TF score, we used the same procedure

    to tackle the ambiguity of uni-gram terms as in the first

    scoring method.

    The second scoring method yielded better results for our

    corpus. The main reason the third one did not performed well

    was that it is biased by the length of the documents. In our

    case, the document (job ad) body can contain detailed job

    functions as well as description of benefits and working

    conditions, resulting in ad lengths ranging from one to

    hundreds of sentences. Given that requirements are usually

    mentioned only once in the ad, the third score gave lower rank

    to important requirements in large ads than in the short ones.

    The requirement ranking was visualized as Wordclouds,

    plotting the phrase font size proportional to its ranking score.

    Wordclouds for mechanical engineering for the three countries

    are presented in Section 4.

    3 MARKET DEMAND BY COUNTRY The job ads obtained from the internet correspond to the

    period June-September 2014, plus January-March 2015, for

    the three countries considered here. It is important to observe

    that different sources were used for obtaining the job ads for

    the three countries, so we cannot conclude with these results

    that there is more demand for engineering majors in one

    country than another. In Fig. 1 we present the Treemap

    corresponding to the Peruvian labor market demand

    considering only engineering majors. There are several

    features worth discussing. For example, it is very common for

    a job ad to include more than one engineering major among its

    requirements. Any prospective job seeker needs to have

    majored in any of the specialties mentioned in the job ad. As a

    consequence, if we add 19502 (Industrial Eng) + 8164

    (Systems Eng) + 7296 (Informatics) + we will obtain 68682, which is larger than the total amount of job ads for

    engineering majors (45368), since many job ads may be

    counted more than once.

    Also, it is important to observe that in Peru and some other

    South American countries, Informatics and Systems

    Engineering are closely related majors and most of the times it

    is difficult to tell them apart. If we consider them as one, then

    we have that mechanical engineering majors are currently the

    fourth most demanded in Peru.

    3 Copyright 2015 by ASME

  • FIGURE 1: Peruvian labor market demand for engineering majors

    In Fig. 2 we present the Treemap corresponding to the Chilean

    labor market demand considering only engineering majors. In

    this case, due to the sources we used to obtain the data, there is

    a significant reduction in the number of job ads analyzed. As a

    consequence, we may expect that this data corresponds to only

    a rough description of the Chilean job market. However, the

    most demanded engineering majors still present a great

    resemblance with the Peruvian case, with a few differences in

    the ranks.

    In Fig. 3 we present the Treemap corresponding to the

    Colombian labor market demand considering again only

    engineering majors. Regarding the top majors, it can be seen

    that the ranking presents a very close resemblance to the

    Peruvian and Chilean cases.

    FIGURE 2: Chilean labor market demand for engineering majors

    4 Copyright 2015 by ASME

  • FIGURE 3: Colombian labor market demand for engineering majors

    As discussed above, it is very common to find more than one

    engineering major as requirement in job ads. This information

    can help us to build a social network kind of graph, in order to visualize the degree of connectedness among these diverse

    technical disciplines. On the left side of Fig. 4, it is shown that

    indeed there is a great degree of relationship among

    engineering fields considering their common mention in job

    ads, for the Peruvian job market case. For example, the two

    most common connections of industrial engineering are

    systems and mechanical engineering. On the right side of Fig.

    4, it is shown the connections that are exclusive to mechanical

    engineering. In this case, the majors that share more job ads

    with mechanical engineering are: industrial (31%), electrical

    (26%) and civil (10%) engineering. Mechanical engineering is

    mentioned exclusively in only 10% of its job ads.

    FIGURE 4: Relationship among Peruvian engineering majors. Left: all connections, right: mechanical engineering connections

    5 Copyright 2015 by ASME

  • FIGURE 5: Relationship among Chilean engineering majors. Left: all connections, right: mechanical engineering connections

    FIGURE 6: Relationship among Colombian engineering majors. Left: all connections, right: mechanical engineering connections

    6 Copyright 2015 by ASME

  • Next, we analyze the Chilean job market case. In Fig. 5, on the

    left side, we find again that there is a great level of

    connectivity among diverse engineering majors. However,

    since the sample for this country is relatively small, many of

    the least frequent connections (corresponding to the thinner

    links) are missing. On the right side of Fig. 5, we observe that

    the most common majors related to mechanical engineering

    are: electrical (33%), industrial (15%) and electronics (13%)

    engineering.

    In the case of the Colombian job market we obtained the

    following results. On the right side of Fig. 6, we also observe a

    great degree of connectivity among diverse majors. On the

    right hand side of Fig. 6, we see that mechanical engineering

    shares most of its job ads with: industrial (23%), electrical

    (22%) and electronics (14%) engineering.

    4 MECHANICAL ENGINEERING REQUIREMENTS During the time intervals considered in our analysis, in

    the Peruvian labor market, mechanical engineering was

    mentioned in 4388 job ads. Almost half of these ads (2141)

    cited maintenance as one of its requirements (see Fig. 7). In

    order to disambiguate the context in which this term is

    employed, it is very useful to look at the related bi- and tri-

    grams. For maintenance we found requirements such as: preventive maintenance, maintenance activities, maintenance management, maintenance plan, among others. Next, we have the requirement control, which can be traced to: quality control, control of projects, process control, among others. In the third place we find processes, which is related to: productive processes, industrial processes, process control, among others.

    In the case of the Chilean labor market, in our sampling, we

    obtained 424 job ads citing mechanical engineering as

    requirement. Among them, almost half (232) mention

    maintenance as a requirement (see Fig. 8). In this case, we find the following related mentions: preventive maintenance, industrial maintenance, maintenance activities, among others. Next, we have control, which is related to quality control, process control, among others. In third place, we find processes, which is traced to: production processes, manufacturing process, quality processes, among others.

    FIGURE 7: Wordcloud for Peruvian mechanical engineering labor market requirements

    7 Copyright 2015 by ASME

  • FIGURE 8: Wordcloud for Chilean mechanical engineering labor market requirements

    Finally, considering the case of the Colombian labor market,

    we obtained that maintenance is once again the most common requirement with 910 job ads, out of a total of 1951

    job ads asking for mechanical engineers (see Fig. 9). This first

    requirement is associated to: preventive maintenance,

    industrial maintenance, equipment maintenance, among others. Following, we find the requirement personal, which is mostly related to interpersonal skills. In third place, we find processes, which is linked to: process control, process quality, industrial process, among others.

    FIGURE 9: Wordcloud for Colombian mechanical engineering labor market requirements

    CONCLUSIONS From our analysis, we have identified that mechanical

    engineering is among the most demanded engineering majors

    in Peru, Chile and Colombia. An interesting feature we have

    unveiled is the friendship between any two majors, defined

    by the amount of job ads shared, which readily permits to

    define something like a social network for majors. In this setting, mechanical engineering appears closely related to

    industrial and electrical engineering majors, among others.

    When analyzing the Wordclouds for mechanical engineering,

    the technical requirement that is the most frequent is

    8 Copyright 2015 by ASME

  • maintenance. Management (project, quality and operations)

    related requirements also rank high, together with a working

    knowledge of English.

    Our results can be used as a complementary tool to what

    educational institutions already employ in order to dimension

    their offer (labor supply) and to define their curriculum

    design, i.e. surveys. It is up to each educational institution to

    define the combination of skills that best supports its

    educational objectives. Some institutions may lean towards

    more technical skills (such as CAD, FEA, etc.), while others

    may give more weight to soft and managerial skills.

    This study represents the use of data analytics as a data-

    driven transparency mechanism to align educational

    institutions to the labor market. By providing this alignment,

    our work can help to ensure conditions for steady economic

    growth, better dimensioned university programs, satisfaction

    for employers and graduates able to work. Overall, our study

    can contribute to improve productivity by helping to provide

    the talented work force South American countries require.

    REFERENCES [1] ManpowerGroup, 2014, The Talent Shortage

    Continues, ManpowerGroup.

    [2] Manyika, J., Remes, J., Woetzel, J., 2014, A

    Productivity Perspective On the Future of Growth, McKinsey

    Quarterly.

    [3] Markes, I., 2006, A Review of Literature on

    Employability Skill Needs in Engineering, European Journal

    of Engineering Education, vol. 31, pp. 637-650.

    [4] Passow, H. J., 2008, What Competencies Should

    Undergraduate Engineering Programs Emphasize? A Dilemma

    of Curricular Design that Practitioners Opinion Can Inform, Ph.D. Dissertation, The University of Michigan.

    [5] Arlett, C., Lamb, F., Dales, R., Willis, L., Hurdle, E.,

    2010, Meeting the Needs of Industry: the Drivers for Change

    in Engineering Education, Engineering Education, vol. 5, pp.

    18-25.

    [6] Male, S. A., Bush, M. B., Chapman, E. S., 2011,

    Australian Study of Generic Competencies Required by

    Engineers, European Journal of Engineering Education, vol.

    36, pp. 151-163.

    [7] Ferguson, C., 2006, Defining the Australian

    Mechanical Engineer, European Journal of Engineering

    Education, vol. 31, pp. 471-485.

    [8] Wolfe, K. E., 2004, Understanding the Careers of the

    Alumni of the MIT Mechanical Engineering Department,

    B.Sc. Dissertation, Massachusetts Institute of Technology.

    [9] Jarosz, J. P., Busch-Vishniac, I. J., A Topical Analysis

    of Mechanical Engineering Curricula, Journal of Engineering

    Education, vol. 95, pp. 241-248.

    [10] Hill, M., Boone, K., Anwar, R., Gaumer, R. B.,

    Camarda K. V., 2014, The Future of Chemical Engineering

    Design: Impact of Faculty Makeup and Industrial Needs,

    Proceedings of the 8th International Conference on

    Foundations of Computer-Aided Process Design - FOCAPD

    2014.

    [11] Xue, Y., 2014, STEM Crisis or STEM Surplus?,

    M.Sc. Dissertation, Massachusetts Institute of Technology.

    [12] Panorama of the Peruvian market demand. Empleatron.com. Empleatron, n.d. Web. 20 April. 2015.

    .

    [13] Relationship among engineering majors in Peru. Empleatron.com. Empleatron, n.d. Web. 20 April. 2015.

    .

    [14] Panorama of the Chilean market demand. Empleatron.com. Empleatron, n.d. Web. 20 April. 2015.

    .

    [15] Relationship among engineering majors in Chile. Empleatron.com. Empleatron, n.d. Web. 20 April. 2015.

    .

    [16] Panorama of the Colombian market demand. Empleatron.com. Empleatron, n.d. Web. 20 April. 2015.

    .

    [17] Relationship among engineering majors in Colombia. Empleatron.com. Empleatron, n.d. Web. 20 April. 2015.

    .

    [18] Rose, S., Engel, D., Cramer, N. and Cowley, W. (2010) Automatic Keyword Extraction from Individual Documents. In M.

    W. Berry & J. Kogan (Eds.), Text Mining: Applications and Theory

    (pp. 11-15), Chichester, UK. [19] Gutwin C., Nevill-Manning C. G., 1999, Domain-

    specific keyphrase extraction, Proceedings of the XVIth

    International Joint Conference on Artificial Intelligence.

    9 Copyright 2015 by ASME