Post-Module
Post-Module Capstone Group Assignment
2022-12-11
Necessay package
1. Data
This dataset contains edges of the talent flow graph. from is the id of the originating firm. to is the id of destination firm. migration_count is the number of workers who have migrated from the from firm to the to firm up until 2018.
df_talent_flows <- read.csv("C://Users//dashv//OneDrive//Desktop//mydata//talent_flows.csv")
head(df_talent_flows)This dataset contains the metadata for each firm. emp_count is the total number of employees LinkedIn lists for the firm. All the other columns are self-explanatory.
df_company <- read.csv("C://Users//dashv//OneDrive//Desktop//mydata//linkedin_company_metadata.csv")
head(df_company)summary(df_company)## company_id name industry city
## Length:473 Length:473 Length:473 Length:473
## Class :character Class :character Class :character Class :character
## Mode :character Mode :character Mode :character Mode :character
##
##
##
##
## country founded hq overview
## Length:473 Min. :1784 Length:473 Length:473
## Class :character 1st Qu.:1911 Class :character Class :character
## Mode :character Median :1962 Mode :character Mode :character
## Mean :1945
## 3rd Qu.:1985
## Max. :2015
## NA's :116
## emp_count
## Min. : 150
## 1st Qu.: 5738
## Median : 13767
## Mean : 36526
## 3rd Qu.: 33361
## Max. :771986
## # create a dataframe called df_edges, where the first two columns are from and to. this will make it easier to work with igraph
df_edges <- df_talent_flows
head(df_edges)2. Get familar with the data
2.1 Create an igraph graph object using the df_edges dataframe
graph_df_talent_flows<- graph_from_data_frame(df_edges, directed = TRUE)
graph_df_talent_flows## IGRAPH 4b2c398 DN-- 473 81114 --
## + attr: name (v/c), migration_count (e/n)
## + edges from 4b2c398 (vertex names):
## [1] at&t ->oracle
## [2] colgate-palmolive ->nike
## [3] agilent-technologies->stryker
## [4] ebay ->expedia
## [5] comcast ->republic-services-inc
## [6] aon ->aig
## [7] costco-wholesale ->apple
## [8] facebook ->cisco
## + ... omitted several edges# How many nodes?
vcount(graph_df_talent_flows)## [1] 473# How many edges?
ecount(graph_df_talent_flows)## [1] 81114# Get nodes
V(graph_df_talent_flows)## + 473/473 vertices, named, from 4b2c398:
## [1] at&t colgate-palmolive
## [3] agilent-technologies ebay
## [5] comcast aon
## [7] costco-wholesale facebook
## [9] john-deere ross-stores
## [11] american-express target
## [13] cme-group jpmorgan-chase
## [15] united-airlines the-home-depot
## [17] xerox wellsfargo
## [19] boeing jefferies
## + ... omitted several vertices# Get edges
E(graph_df_talent_flows)## + 81114/81114 edges from 4b2c398 (vertex names):
## [1] at&t ->oracle
## [2] colgate-palmolive ->nike
## [3] agilent-technologies->stryker
## [4] ebay ->expedia
## [5] comcast ->republic-services-inc
## [6] aon ->aig
## [7] costco-wholesale ->apple
## [8] facebook ->cisco
## [9] john-deere ->ge
## [10] ross-stores ->walmart
## + ... omitted several edges2.2 Calculate the in-degree and out-degree for each firm. What are the top 10 firms with the highest in-degree? What are the 10 firms with the highest out-degree? Describe in your own words what these metrics mean.
# Get a list of all possible company by looking at the
# 'from' and 'to' columns
company <- unique(c(df_edges[, 1], df_edges[, 2]))
# Count how many inbound links all company have
in_count <- table(factor(df_edges[, 2], levels = company))
# Turn these counted objects into dataframes
in_count <- data.frame(in_count)
colnames(in_count) <- c("company", "freq")
df_in_count <- data.frame(in_count)
titles <- read.csv("C://Users//dashv//OneDrive//Desktop//mydata//linkedin_company_metadata.csv")
join_df_in_count <-left_join(df_in_count, titles, by = c(company = "company_id"))
top_df_in_count <- join_df_in_count %>%
arrange(desc(freq)) %>%
head(10)
join_df_in_countprint("top 10 firms with highest in-degree")## [1] "top 10 firms with highest in-degree"top_df_in_count$name## [1] "IBM" "Accenture"
## [3] "Hewlett Packard Enterprise" "AT&T"
## [5] "Bank of America" "Amazon"
## [7] "Wells Fargo" "JPMorgan Chase & Co."
## [9] "Microsoft" "Citi"#---------------------------------------------------------------------------------------------------------------------
# Count how many outbound links all company have
out_count <- table(factor(df_edges[, 1], levels = company))
# Turn these counted objects into dataframes
out_count <- data.frame(out_count)
colnames(out_count) <- c("company", "freq")
df_out_count <- data.frame(out_count)
join_df_out_count <-left_join(df_out_count, titles, by = c(company = "company_id"))
top_df_out_count <- join_df_out_count %>%
arrange(desc(freq)) %>%
head(10)
print("top 10 firms with highest out-degree")## [1] "top 10 firms with highest out-degree"top_df_out_count$name## [1] "IBM" "AT&T"
## [3] "Hewlett Packard Enterprise" "JPMorgan Chase & Co."
## [5] "Bank of America" "Accenture"
## [7] "GE" "Wells Fargo"
## [9] "Citi" "Target"A: From the result, IBM has the most # of talents in/out of the company which means that it has the highest rotation rate among others.
2.3 You’ll notice that the firms with the highest degree are biased towards larger firms. Explain why we might expect this type of correlation?
Answer: Larger companies tend to have higher reputation and attract more talents from the markets. Also, talents from larger companies will have more opportunities in the job markets. Both reasons cause large companies to have highest degree.
2.5. Report and interpret the results of the regression (What do the coefficients mean? What do the p-values mean?)
Summary Both model have statistically significant result on coefficient. For every one unit change in employee count of the company, there in-degree will increase by 9.545e-04 units. While, for every one unit change in employee count of the company, the out-degree will increase by 1.055e-03 units.
3. Weigted Edges
3.1. Calculate the weight for each edge, and add this as a new column called weight to the existing df_edges dataframe.
The weight is migration count (migration_count) divided by the total number of employees (emp_count) of the starting firm.
# Add the weight column to df_edges
df_edges <- df_edges %>%
inner_join(df_company, by = c("from" = "company_id")) %>%
select(from, to, migration_count, emp_count) %>%
mutate(weight = migration_count/emp_count)3.2. Create a directed graph (with weighted edges) using the df_edges dataframe.
# select the "from", "to", and "weight" columns
df_weighted_edges <- df_edges %>%
select(from, to, weight)
# create the igraph
graph_df_talent_flows_weighted <- graph_from_data_frame(df_weighted_edges,directed = TRUE)
# view vertex
V(graph_df_talent_flows_weighted)## + 473/473 vertices, named, from 4ba0ad3:
## [1] at&t colgate-palmolive
## [3] agilent-technologies ebay
## [5] comcast aon
## [7] costco-wholesale facebook
## [9] john-deere ross-stores
## [11] american-express target
## [13] cme-group jpmorgan-chase
## [15] united-airlines the-home-depot
## [17] xerox wellsfargo
## [19] boeing jefferies
## + ... omitted several vertices# view edges
E(graph_df_talent_flows_weighted)## + 81114/81114 edges from 4ba0ad3 (vertex names):
## [1] at&t ->oracle
## [2] colgate-palmolive ->nike
## [3] agilent-technologies->stryker
## [4] ebay ->expedia
## [5] comcast ->republic-services-inc
## [6] aon ->aig
## [7] costco-wholesale ->apple
## [8] facebook ->cisco
## [9] john-deere ->ge
## [10] ross-stores ->walmart
## + ... omitted several edges3.3. Take the top 10 edges with the largest weight in the graph (and the associated nodes), and visualize it by plotting the graph.
# find the top 10 edges by weight and make it as a new data frame
df_weighted_edges_top10 <- df_weighted_edges %>%
arrange(desc(weight))
df_weighted_edges_top10 <- as.data.frame(head(df_weighted_edges_top10, n=10))
print(df_weighted_edges_top10)## from to weight
## 1 hp hewlett-packard-enterprise 0.13910068
## 2 abbott- abbvie 0.05724070
## 3 allegion-us ingersoll-rand 0.05494505
## 4 verisign symantec 0.05486244
## 5 agilent-technologies keysight-technologies 0.04563948
## 6 conocophillips phillips66co 0.03947543
## 7 juniper-networks cisco 0.03780439
## 8 ebay paypal 0.03577366
## 9 weyerhaeuser international-paper 0.03391256
## 10 host-hotels-&-resorts marriott-international 0.03214286# create igraph for the top10 weighted edges
graph_df_talent_flows_weighted_top10 <- graph_from_data_frame(df_weighted_edges_top10, directed=TRUE)
# plot graph
plot.igraph(graph_df_talent_flows_weighted_top10,
layout = layout.kamada.kawai,
vertex.color = "plum2",
vertex.size = 12,
vertex.label.color = "black",
vertex.label.font = 2,
vertex.label.cex = .7,
edge.color = "slategrey",
edge.curved = .1,
edge.arrow.size = .6,
edge.width = 1,
edge.label = round(E(graph_df_talent_flows_weighted_top10)$weight,2),
edge.label.font = 2,
edge.label.cex = .8,
edge.label.color = "blue",
asp = -.2,
margin = -.01,
main = "The top 10 edges with the largest weight")
3.4. Interpret your graph. What does it mean if an edge has a high weight? Do you notice anything interesting in the graph?
Answer: The higher the weight, the larger the number of employees from company A jumping to company B. Looking at the graph above, the highest weight 0.14 belongs to “hp –> hp-enterprise”, which indicates that there is a high movement between the subsidiaries of the same Hewlett Packard group. Talent flow within the same group or between affiliated companies is more frequent.Ebay and Paypal is another interesting example, as Ebay had acquired Paypal in 2015.It is also interesting to note that majority of talent flows in and out within the same sector, symantec and verisign in cybersecurity, keysight and agilent in electronics sector, marriot and host-hotel & resorts etc.
4. PageRank
4.1 Explain the intuition behind PageRank using the random surfer model. How is the random surfer model different for graphs with weighted edges?
Answer: PageRank is a metric to rank nodes within a network based on their importance. The random surfer model was essentially first applied to the web network. It assumes that the “random surfer” is someone who starts on a random webpage(node in this case), and then keeps clicking through the links within the first page and subsequent webpages for an infinitely long time. As the surfer does this for an infinitely long time, it will end up spending most amount of time on a webpage that has the highest number of in-bound links, thus highlighting the importance/centrality of the webpage. Similarly, when you extend this concept to a network, PageRank can be used to rank nodes within a network based on how frequently a surfer randomly traversing the network is expected to end up at the most important/central nodes.
For a network with weighted edges, the probability of a random surfer clicking through from a webpage to another is governed by the weight of the edge. An edge with a higher weight has a higher probability of getting clicked through. In case of unweighted edges, the probability is distributed evenly across all edges.
4.2 Calculate the weighted PageRank and the unweighted PageRank for all the nodes in the graph. What are the 10 nodes with the highest unweighted PageRank? What are the 10 nodes with the highest weighted PageRank?
# find the top 10 edges by weight and make it as a new data frame
df_pagerank_unweighted <- as.data.frame(page_rank(graph_df_talent_flows,vids = V(graph_df_talent_flows),directed = TRUE,weights = NULL)$vector)
colnames(df_pagerank_unweighted) <- c("Page_Rank")
df_pagerank_unweighted_top10 <- df_pagerank_unweighted %>%
arrange(desc(Page_Rank))
df_pagerank_unweighted_top10 <- as.data.frame(head(df_pagerank_unweighted_top10, n=10))
print("top 10 highest ranked companies based on pagerank(unweighted)")## [1] "top 10 highest ranked companies based on pagerank(unweighted)"print(df_pagerank_unweighted_top10)## Page_Rank
## wellsfargo 0.005052306
## ibm 0.005042449
## accenture 0.004941054
## bank-of-america 0.004872047
## at&t 0.004823967
## hewlett-packard-enterprise 0.004822008
## amazon 0.004773077
## jpmorgan-chase 0.004701858
## microsoft 0.004629907
## citi 0.004615667df_pagerank_weighted <- as.data.frame(page_rank(graph_df_talent_flows_weighted,vids = V(graph_df_talent_flows_weighted),directed = TRUE,weights = graph_df_talent_flows_weighted$weight)$vector)
colnames(df_pagerank_weighted) <- c("Page_Rank")
df_pagerank_weighted_top10 <- df_pagerank_weighted %>%
arrange(desc(Page_Rank))
df_pagerank_weighted_top10 <- as.data.frame(head(df_pagerank_weighted_top10, n=10))
print("top 10 highest ranked companies based on pagerank(weighted)")## [1] "top 10 highest ranked companies based on pagerank(weighted)"print(df_pagerank_weighted_top10)## Page_Rank
## ibm 0.02584084
## microsoft 0.02254956
## hewlett-packard-enterprise 0.02095859
## wellsfargo 0.02031841
## bank-of-america 0.01918484
## jpmorgan-chase 0.01867777
## citi 0.01614856
## accenture 0.01609669
## google 0.01494275
## oracle 0.013327364.2 Plot the distribution(histogram) of weighted and unweighted PageRanks. Comment on the differences, and explain why these might be different.
ggplot(df_pagerank_unweighted,aes(x=Page_Rank)) + geom_histogram(fill = "slategrey") +labs(title = "Histogram for PageRank(Unweighted)", x = "Page_Rank", y = "Frequency")
ggplot(df_pagerank_weighted,aes(x=Page_Rank)) + geom_histogram(fill = "slategrey") +labs(title = "Histogram for PageRank(Weighted)", x = "Page_Rank", y = "Frequency")
Answer: For unweighted page-rank, the distribution is
much more uniform and it tells us that the migration between companies
happens across a wide breadth of companies and there are lot of
companies that individuals move to. However, upon looking at the
weighted page-rank, it is fairly clear that the larger companies stand
out in terms of the % of individuals moving to them as a function of
their overall size. The page_rank for the top 10 companies when
considering weights appears to be significantly higher than that of the
smaller companies, and hence the distribution appears to be skewed
towards the left.
5. Community Detection
5.1 . Walktrap is an algorithm for community detection that is computationally very efficient and that we discussed briefly in class. Use the walktrap.community command in igraph to detect communities of firms. You can play around with steps argument to get different communities, the default steps=4 seems to produce good results.
library(clustAnalytics)
walktrap <- walktrap.community(graph_df_talent_flows_weighted, steps = 4)
modularity(walktrap)## [1] 0.3920425plot(walktrap, graph_df_talent_flows_weighted,
vertex.color=membership(walktrap),
layout = layout_with_fr,
vertex.size = 3,
vertex.label = NA,
edge.label = NA,
edge.color = "lightgrey",
edge.curved = .1,
edge.arrow.size = .1,
edge.width = .5,
asp = -.2,
main = "Communities of Firms")
#### 5.2 . Inspect the members of each community. What commonalities do
you observe within each community?
Answer: Firms belonging to the same industry have a tendency to be in the same community, which can be expected given that people will look for jobs that fit their skills and experiences and majority of job changes tend to be within the same industry. For instance, financial institutions such as Wells Fargo, JP Morgan Chase, Citi, US Bank and PNC Bank all belong to the same community. Similar trends can be observed across other industries as well.
membership(walktrap)## at&t colgate-palmolive
## 2 4
## agilent-technologies ebay
## 1 1
## comcast aon
## 2 7
## costco-wholesale facebook
## 2 1
## john-deere ross-stores
## 4 2
## american-express target
## 2 2
## cme-group jpmorgan-chase
## 7 7
## united-airlines the-home-depot
## 2 2
## xerox wellsfargo
## 2 7
## boeing jefferies
## 2 7
## cardinal-health cisco
## 4 1
## ibm nordstrom
## 2 2
## johnson-controls honeywell
## 4 4
## unitedhealth-group exxonmobil
## 7 5
## microsoft netapp
## 1 1
## state-street eaton-corporation
## 7 4
## church-&-dwight-co-inc general-dynamics
## 4 2
## gap-inc- alliant-energy
## 2 5
## intuit ppg-industries
## 1 4
## kohls-department-stores cintas
## 2 2
## apple charles-schwab
## 1 7
## procter-&-gamble pepsico
## 4 4
## stryker analog-devices
## 6 1
## mckesson motorolasolutions
## 2 1
## fedex spglobal
## 2 2
## leggett-&-platt the-hershey-company
## 2 4
## applied-materials medtronic
## 1 6
## ups johnson-&-johnson
## 2 6
## autodesk hca
## 1 7
## 3m travelers
## 4 7
## zions-bancorporation broadcom
## 7 1
## abbott- hewlett-packard-enterprise
## 6 1
## chevron pvh
## 3 2
## kellogg-company bank-of-america
## 4 7
## citi walgreens
## 7 2
## newellbrands dish-network
## 4 2
## mattel us-bank
## 4 7
## intel-corporation hilton-worldwide
## 1 2
## goldman-sachs parker-hannifin
## 7 4
## cummins-inc dow-chemical
## 4 5
## metlife american-water
## 7 2
## lowe%27s-home-improvement yum-brands
## 2 2
## general-motors csx-transportation
## 4 2
## raytheon marathon-oil-corporation
## 2 3
## walmart ford-motor-company
## 2 4
## discover-financial-services mohawk-industries
## 7 2
## fox-filmed-entertainment autozone
## 2 2
## citrix concho-resources
## 1 3
## visa american-electric-power
## 1 5
## cvs-health the-kraft-heinz-company
## 2 4
## ameren tractor-supply-company
## 5 2
## aimco pnc-bank
## 9 7
## h&r-block amazon
## 2 1
## mastercard schlumberger
## 2 3
## adm the-walt-disney-company
## 4 2
## thermo-fisher-scientific eversourceenergy
## 6 5
## ecolab corning-incorporated
## 4 4
## sherwin-williams host-hotels-&-resorts
## 2 2
## w.w.-grainger arthur-j--gallagher-and-co
## 2 7
## paychex the-estee-lauder-companies-inc-
## 2 4
## jacobs caterpillar-inc
## 5 4
## insidepmi idexx-laboratories
## 4 4
## textron willis-towers-watson
## 4 7
## tiffany-and-co conocophillips
## 2 3
## darden-restaurants charter-communications
## 2 2
## cboe illumina
## 7 6
## symantec alcoa
## 1 4
## fastenal borgwarner
## 2 4
## general-mills xilinx
## 4 1
## unionpacific pfizer
## 2 6
## cognizant nasdaq-omx
## 2 2
## ipg-photonics iron-mountain
## 6 2
## vertex-pharmaceuticals huntington-national-bank
## 6 7
## merck the-j-m--smucker-company
## 6 4
## ge avery-dennison
## 4 4
## principal-financial-group flowserve
## 7 5
## consumers-energy baker-hughes
## 5 3
## nike jb-hunt-transport-services-inc
## 4 2
## google exelon
## 1 5
## salesforce keysight-technologies
## 1 1
## robert-half-international sysco
## 2 4
## maxim-integrated tsys
## 1 2
## davita phillips66co
## 2 3
## macy norwegian-cruise-line
## 2 10
## labcorp dovercorp
## 4 4
## twitter the-coca-cola-company
## 1 4
## marriott-international allergan
## 2 6
## simon-property-group tjx
## 8 2
## air-products gartner
## 5 1
## people%27s-united-bank halliburton
## 7 3
## abbvie dominion
## 6 5
## equifax perrigo
## 2 6
## incyte northern-trust
## 6 7
## american-tower equity-residential
## 2 9
## campbell-soup-company western-digital
## 4 1
## brown-forman hess-corporation
## 4 3
## morgan-stanley c-h-robinson
## 7 2
## foot-locker the-hartford
## 2 7
## con-edison skyworks-solutions-inc
## 5 1
## bd1 firstenergy-corp
## 6 5
## boston-properties perkinelmer
## 8 6
## whirlpool-corporation pentair
## 4 4
## electronic-arts cbs-com
## 1 2
## cimarex-energy zimmerbiomet
## 3 6
## fis amgen
## 7 6
## international-paper texas-instruments
## 4 1
## the-clorox-company zoetis
## 4 6
## adobe genuine-parts-company
## 1 2
## fifth-third-bank regions-financial-corporation
## 7 7
## atmos-energy waste-management
## 5 2
## biogen- seagate-technology
## 6 1
## the-linde-group red-hat
## 5 1
## northrop-grumman-corporation micron-technology
## 2 1
## ingersoll-rand progressive-insurance
## 4 7
## freeport-mcmoran-inc verizon
## 5 2
## constellation-brands lyondell-basell
## 4 5
## regency-centers starbucks
## 8 2
## danaher celgene
## 4 6
## public-storage fmc-corporation
## 2 5
## nvidia harley-davidson-motor-company
## 1 4
## oxy carmax
## 3 2
## valero-energy-corporation accenture
## 5 2
## newscorp mylan
## 1 6
## fluor allstate
## 5 7
## prudential-financial fiserv
## 7 7
## newmont-mining-corporation nielsen
## 5 4
## chipotle-mexican-grill aes
## 2 5
## aig gilead-sciences
## 7 6
## capital-one bristol-myers-squibb
## 7 6
## best-buy henry-schein
## 2 4
## qualcomm oracle
## 1 1
## dollar-tree-stores pca
## 2 4
## dr-horton raymond-james-financial-inc-
## 2 7
## vulcan-materials-company cbre
## 5 8
## regeneron-pharmaceuticals united-technologies
## 6 4
## emerson eli-lilly-and-company
## 4 6
## national-oilwell-varco franklin-templeton-investments
## 3 7
## ameriprise-financial-services-inc lockheed-martin
## 7 2
## cigna amd
## 7 1
## moodys-corporation stanley-black-decker-inc
## 7 4
## nektar-therapeutics ansys-inc
## 6 1
## pioneer-natural-resources-company kinder-morgan
## 3 3
## mcdonald%27s-corporation citizens-bank
## 2 7
## ulta humana
## 2 7
## suntrust-bank nucor-corporation
## 7 5
## uhs juniper-networks
## 7 1
## helmerich-&-payne viacom
## 3 2
## m&t-bank kroger
## 7 2
## hp lam-research
## 1 1
## mgm-resorts-international adp
## 2 2
## williams-company duke-energy-corporation
## 3 5
## westrockcompany rollins-inc.
## 4 2
## devon-energy cadence-design-systems
## 3 1
## western-union t--rowe-price
## 2 7
## f5-networks cerner-corporation
## 1 2
## berkshire-hathaway aflac
## 7 2
## lennar torchmark-corporation
## 2 7
## dte-energy advance-auto-parts
## 5 2
## activision under-armour
## 1 2
## global-payments mondelezinternational
## 7 4
## hasbro chubb
## 4 7
## priceline-com paccar
## 2 2
## ball tyson-foods
## 4 2
## royal-caribbean masco-corporation
## 10 4
## hanesbrands-inc- kimberly-clark
## 4 4
## ims-health ihs
## 6 5
## garmin-international alliance-data
## 1 7
## intuitive-surgical amerisourcebergen
## 6 4
## the-cincinnati-insurance-companies assurant
## 7 7
## limited-brands bny-mellon
## 2 7
## hologic southern-company
## 6 5
## xcel-energy lincoln-financial-group
## 5 7
## marathon-petroleum-company wynn-las-vegas
## 5 2
## kansas-city-southern-railway verisign
## 2 1
## american-airlines pultegroup
## 2 2
## southwest-airlines anadarko-petroleum-corporation
## 2 3
## eastman-chemical-company mccormick
## 5 4
## coty teleflex
## 4 6
## iff sealed-air-corporation
## 4 4
## kla-tencor blackrock
## 1 7
## prologis rockwell-automation
## 8 4
## te-connectivity discovery-communications
## 4 2
## dentsplysirona waters
## 6 6
## equinix centurylink
## 1 2
## paypal vf-corporation
## 1 4
## carnival-cruise-lines svb-financial-group
## 10 1
## itw norfolk-southern
## 4 2
## avalonbay-communities alexion-pharmaceuticals
## 9 6
## hollyfrontier-corporation xylem-inc-
## 3 4
## celanese synchrony-financial
## 5 7
## sempra-energy macerich
## 2 8
## ralph-lauren nrg-energy
## 2 5
## united-rentals ametek
## 2 4
## boston-scientific akamai-technologies
## 6 1
## etrade alaska-airlines
## 7 2
## martin-marietta-materials nextera-energy-resources
## 5 5
## mid-america-apartment-communities harris-corporation
## 8 2
## marsh-&-mclennan-companies-inc- bb&t
## 7 7
## republic-services-inc altria
## 2 4
## edwards-lifesciences delta-air-lines
## 6 2
## ipg entergy
## 2 5
## keybank broadridge-financial-solutions
## 7 7
## microchip-technology baxter-healthcare
## 1 6
## anthem vornado-realty-trust
## 7 8
## netflix molson-coors
## 1 4
## expeditors dollar-general
## 2 2
## flir-systems michael-kors
## 1 2
## expedia welltower
## 1 8
## verisk-analytics align-technology
## 7 1
## jack-henry-&-associates essex-property-trust
## 2 9
## pseg fleetcor
## 5 2
## wellcare ppl-corporation
## 7 5
## cabot-oil-&-gas invesco-ltd
## 3 7
## digitalrealty resmed
## 8 6
## hormel-foods allegion-us
## 4 4
## extra-space-storage o%27reilly-auto-parts
## 2 2
## crown-castle weyerhaeuser
## 2 4
## nisource unum
## 5 7
## mosaiccompany centerpoint-energy
## 5 5
## msci-inc centene-corporation
## 7 7
## take-2-interactive-software-inc- tripadvisor
## 1 1
## comerica-bank snap-on-tools
## 7 4
## varian-medical-systems copart
## 1 2
## duke-realty-corporation abiomed
## 8 6
## oneok qorvo
## 3 1
## albemarle lkq-corporation
## 5 2
## noble-energy udr
## 3 9
## cf-industries realty-income-corporation
## 5 1
## fortinet first-republic-bank
## 1 7
## edison-international hcp
## 5 8
## sba-communications wec-energy-group
## 2 5
## intercontinentalexchange-inc- diamondback-energy-services
## 7 3
## alexandria-real-estate-equities-inc- everest-reinsurance
## 8 7
## kimco-realty-corporation apache-corporation
## 8 3
## eog-resources monster-energy_2
## 3 4
## sl-green federal-realty-investment-trust
## 8 8
## ventas-inc.
## 86.
6.1. In your own words explain what assortative mixing means.
Answer: Assortativity is a bias in favor of connections between network nodes with similar characteristics and behavior. That is to say, the more similar the entities are, the more likely the entities will interact. In class, smokers were given as an example of shared behavior which leads to Assortativeity mixing. Another example with high school grades, is higher performing students are more likely to share similar behavior with other high performing students. Similarly, low performing students are more likely to share similar behavior with low performing students. This may be due to ability and being placed in the same classes. This leads to more interaction and stronger relationships.
6.2. Calculate the level of assortative mixing for industry, and interpret your results. HINT: Use the assortativity command in igraph. The types1 parameter should be a vector of industries corresponding to the nodes (firms).
Answer: Using assortativity(), we get the following assortativity mixing for industry:
assortativity(graph_df_talent_flows_weighted , V(graph_df_talent_flows_weighted), types2 = NULL, directed = TRUE)## [1] -0.060634156.3. What are some other characteristics (come up with 2) along which you may find assortative mixing in the talent flow graph. What about disassortative mixing (come up with 1)? This is an open-ended question and the characteristics don’t have to be in the dataset but make sure to explain your answers.
Answer: Some of the characteristics, assortative mixing in the talent flow graphs may include industry(line of business), and company size. This can be explained because people are creatures of habits. As an example, person who is intimate with the advertising industry, may want to move to another company who works in advertising. From the company perspective, the company may only want to hire someone who understands the industry, because they believe the new hire may perform better. As for company size, people who changes jobs, may look for companies that have similar size. For example, someone who works in a 10k+ persons company may want to remain in a large company, because if the person moves to a start-up the person may be asked to wear multiple hats. This person may not want that to happen, therefore, sticks to similar size companies.
A characteristic of dissortative mixing may be company financial/ownership status. One way to describe it would be if a company gets acquired by another company. This is also a way personnel move from one company to another.