December 2022 - Migration - mailman.cs.vt.edu

THANKS - VT Migration Prediction Update
by Flynn A Rapavi 27 Dec '22

27 Dec '22

Brian, Sincere thanks for the update- good news on the ingest rate for the second source. A lot of progress made, to date- awesome. Hope you had a great holiday. Talk with you soon. V/r, Alex [Office of the Director of National Intelligence Seal] Alexander Rapavi PMM, GTOC/IN/CN NIM-WHT Office of the Director of National Intelligence 703.275.2711 From: Brian Mayer <bmayer(a)cs.vt.edu> Sent: Saturday, December 24, 2022 9:49 PM To: Flynn A Rapavi <FLYNNAR(a)dni.gov>; TIMOTHJG <TIMOTHJG(a)nctc.gov>; Nicole C Peruyera <NICOLECP(a)nctc.gov>; Deborah A Bowman <DEBORAB2(a)dni.gov>; HOLLYTP <HOLLYTP(a)ucia.gov> Cc: migration(a)cs.vt.edu Subject: VT Migration Prediction Update All, Please find attached our biweekly status report. Let us know if there is anything you would like to discuss further. Have a happy holiday. Thanks, Brian (and team) -- Brian Mayer bmayer(a)cs.vt.edu<mailto:bmayer@cs.vt.edu> 540-231-5907 Sanghani Center for Artificial Intelligence and Data Analytics<https://sanghani.cs.vt.edu/> - Virginia Tech<https://vt.edu/>

1 0

VT Migration Prediction Update
by Brian Mayer 24 Dec '22

24 Dec '22

All, Please find attached our biweekly status report. Let us know if there is anything you would like to discuss further. Have a happy holiday. Thanks, Brian (and team) -- Brian Mayer bmayer(a)cs.vt.edu 540-231-5907 Sanghani Center for Artificial Intelligence and Data Analytics <https://sanghani.cs.vt.edu/> - Virginia Tech <https://vt.edu/>

1 0

Fwd: Status Request
by Brian Mayer 14 Dec '22

14 Dec '22

Some key takeaways 1. Monthly data is based on the date of encounter (not entry or expulsion). 2. Encounter data stands even if the migrant is admitted or expelled. I also notice that on the page for the "daily" data ( https://www.cbp.gov/document/stats/us-border-patrol-nationwide-apprehension…) says "Apprehensions" so I assume these data don't include Title 8 Inadmissibles (in addition to not including Title 42). Therefore, I think it might make sense to try and predict Title 8 Inadmissibles (attempted legal entry) and Title 8 Apprehensions (attempted "illegal" entry) separately so we can better match to the daily data. I also noted (in the monthly data and the monthly data dictionary <https://www.cbp.gov/document/stats/nationwide-encounters>) that Title 8 Inadmissables are encountered by OFO and Title 8 Apprehensions are encountered by CBP. Similarly, Title 42 Expulsions are encountered by OFO, and Title 42 Apprehensions are encountered by CBP. So the difference is probably similar (Title 42 Expulsions are legal attempts and Title 42 Apprehensions are "illegal" attempts) So this means there will be a slight change to the aggregation I proposed in the other email string. The new aggregation will be: Monthly data (Oct 2019 - Oct 2022): *<month/year, country of origin, title of authority, component, quantity>* Daily data (Oct 2008 - Sep 2021): *(we probably only need to go back to Oct 2019 which is the same as the monthly data and 6 months before Title 42, maybe a little earlier)* *<date, country of origin, quantity> **(Only represents Title 8 Apprehensions**)* And we should forecast for each grouping separately. The monthly data has 4 groups per country (2 titles x 2 components). The daily data has 1 group per country (Title 8 Apprehensions, CBP). Depending on what we see we may want to combine Title 8 and Title 42 as there may be a more consistent (or forecasteable) stream since there really doesn't seem to be much reason to classify someone as Title 8 or 42. Even if we do this I am sure we will still see jumps in mar 2020 and Dec 2022. ---------- Forwarded message --------- From: Flynn A Rapavi <FLYNNAR(a)dni.gov> Date: Wed, Dec 14, 2022 at 7:40 PM Subject: RE: Status Request To: Brian Mayer <bmayer(a)cs.vt.edu> Cc: Naren Ramakrishnan <naren(a)cs.vt.edu>, Nicole C Peruyera < NICOLECP(a)nctc.gov>, Flynn A Rapavi <FLYNNAR(a)dni.gov> Brian, Thank you for the response. I completely understand- I have been there! Regarding the rescission of Title 42, I’ve attached a few documents from the Congressional Research Service that might be of use (good summaries anyway) as well as provided some additional information/insights. Title 42 provided authority under the US public health code that supported expulsions (pursuant to CDC, ‘Public Health Reassessment and Order Suspending the Right to Introduce Certain Persons from Countries Where a Quarantinable Disease Exists,’ reference 86 Federal Register 42828-42841, August 5, 2021). When CBP encounters a migrant entering the US illegally between established ports of entry, those migrants are processed for removal under Title 8 (immigration law) and then asked whether they fear persecution or intend to seek asylum. Title 42 was put in place in CY 2020; provisions within Title 42 mandated CBP prohibit the entry of specific migrants/migrant groups to help prevent the spread of COVID-19 in border facilities. This in turn allowed removal of these individuals/groups generally to Mexico – under the Title 42 provisions, CBP would not necessarily offer the standard fear/asylum seeking interview normally required/conducted under Title 8. The non-incremental rescission of Title 42 is providing a sound impetus/incentive for many migrants; it’s already creating substantial upticks in daily encounters as you are likely seeing in the data A good source for data on this is (you are probably already familiar with this): https://www.cbp.gov/newsroom/stats/cbp-enforcement-statistics/title-8-and-t… Regarding the scenarios referenced below, you are indeed correct. Those values are associated with *initial encounters* vice entrance/final admittance figures. Expulsions (and/or readmissions) are treated as separate metrics, as well - to include those expulsions not invoking Title 42. As I understand the counting system, the encounter data stands even if the migrant is admitted or expelled. So effectively, a migrant can be in effect counted twice, albeit in different categories (categorical severability). The passage below from CBP indicates *inadmissibles* are those migrants encountered at established Ports of Entry but who are seeking lawful admission to the US; *Apprehension *refers to the physical control or temporary detainment of a person who is not lawfully in the US (which to me sounds like forced expulsion of migrants who crossed the border either via a POE (it does occur) or between POEs, and is subsequently caught and likely does not pass required screening questions. If caught engaging in criminal act, they would be arrested and likely formally deported). I assume deportations are a separate category altogether. Regarding your second reference with respect to ENTRY, APPR_DT_TIME and FINAL BOOKOUT, I believe Title 42 expulsions are treated and recorded separately – and I believe your characterizations of the labels are accurate. While it comports with what I’ve read and with conversations I’ve had, to be certain, I will contact CBP and get specifics on this. In short, I want to make sure information I’ve been provided in the past has not changed- I will get back to you ASAP—and in the meantime, I am hopeful what I was able to provide helps in the interim- Best regards, Alexander R. *Nationwide Enforcement Encounters: Title 8 Enforcement Actions and Title 42 Expulsions Fiscal Year 2023* On March 21, 2020 the President, in accordance with Title 42 of the United States Code Section 265, determined that by reason of existence of COVID-19 in Mexico and Canada, there is a serious danger of the further introduction of COVID-19 into the United States; that prohibition on the introduction of persons or property, in whole or in part, from Mexico and Canada is required in the interest of public health. Under this order, CBP is prohibiting the entry of certain persons who potentially pose a health risk, either by virtue of being subject to previously announced travel restrictions or because they unlawfully entered the country to bypass health screening measures. To help prevent the introduction of COVID-19 into border facilities and into the United States, persons subject to the order will not be held in congregate areas for processing and instead will immediately be expelled to their country of last transit. In the event a person cannot be returned to the country of last transit, CBP works with interagency partners to secure expulsion to the person’s country of origin and hold the person for the shortest time possible. This order does not apply to persons who should be excepted based on considerations of law enforcement, officer and public safety, humanitarian, or public health interests. Expulsions under Title 42 are not based on immigration status and are tracked separately from immigration enforcement actions, such as apprehension or inadmissibility, that are regularly reported by CBP. To access the data used to build this dashboard, please visit the CBP Data Portal. https://www.cbp.gov/document/stats/southwest-land-border-encounters *From:* Brian Mayer <bmayer(a)cs.vt.edu> *Sent:* Tuesday, December 13, 2022 9:34 PM *To:* Flynn A Rapavi <FLYNNAR(a)dni.gov> *Cc:* Naren Ramakrishnan <naren(a)cs.vt.edu>; Nicole C Peruyera < NICOLECP(a)nctc.gov> *Subject:* Re: Status Request Alex, Unfortunately, we are still tuning the models and the forecasts are not yet calibrated for reliability. Since we are still tuning the models there is both data and model drift that we will need to take into account, but plan to get to it soon. That being said we were curious if we could get some more insight into the impacts of the Titles of Authority so we can be sure to capture this appropriately moving forward. In the CBP monthly data (here <https://www.cbp.gov/document/stats/nationwide-encounters>) there are three different scenarios for Title of Authority and Encounter Type (Title 8: Inadmissibles, Title 8: Apprehension, and Title 42: Expulsions). We just wanted to confirm that these numbers record * when the person is encountered, not when they enter*, and not when they are returned to another country. Are they also still counted in these data if they are accepted in the US after the encounter? Can you tell us the difference between Title 8: Inadmissibles and Title 8: Apprehension? In the more detailed data (here <https://www.cbp.gov/document/stats/us-border-patrol-nationwide-apprehension…>) only the Title 8's are reported so *we assume* *that *Title 42 expulsions aren't recorded in this data set. Each record has an ENTRY_DT (*assumed to be* the known or estimated DateTime of the person's entry into the US), APPR_DT_TIME (*assumed to be* the DateTime of the encounter), FINAL BOOKOUT DATE (*assumed to be* the date the case is resolved). Are these assumptions correct? And again are they still counted in these data if they are accepted in the US after the encounter? I think some of these questions get at similar topics that we would have covered in the call that was scheduled pre-Thanksgiving. I would be happy to jump on a call this week if that helps. I am pretty flexible the rest of the week. Thanks, Brian On Mon, Dec 12, 2022 at 11:33 AM Flynn A Rapavi <FLYNNAR(a)dni.gov> wrote: Brian, Definitely appreciate the specifics. Data volume is impressive and will likely yield some beneficial insights- decidedly excited about the output. As you can imagine, with pending changes to specific provisions in US immigration law (e.g. starting 21 Dec timeframe), we’ve been getting some questions (from internal, external sources) regarding possible projections in the lead-up. Do you have the ability to execute some prelim runs based on existing data set availability (e.g. prior to 21 Dec)? Would said runs (even if limited in duration) be executable this week? Recognize the limitations in doing this, but would be beneficial for us to inform conversations and for comparative purposes. Regarding your questions, I’ve reached out to my contacts twice and have not heard back. I will call them this week and see if I can lock down whether data post 4Q21 can be made available and also if they can do a direct push of the executive leadership reports- agree, getting the raw data would be ideal. Will see what I can do— Thanks for the big forward-lean on this project— V/R, Alexander R. PMM/TOC/CD DMI/NIM-WHT 703.275.2711 *From:* Brian Mayer <bmayer(a)cs.vt.edu> *Sent:* Thursday, December 08, 2022 8:06 PM *To:* Flynn A Rapavi <FLYNNAR(a)dni.gov>; Naren Ramakrishnan <naren(a)cs.vt.edu> *Cc:* Nicole C Peruyera <NICOLECP(a)nctc.gov> *Subject:* Re: Status Request Alex, We have already ingested 2.8M articles from Oct 1, 2020 - current from the one source and we are ingesting 3000+ articles per day coming from over 4000 potential domains. We are finishing up the enrichment of these articles to prepare them to be used by the forecasting model. This involves extracting entities and creating keyword counts. We should be ready to start giving forecasts by the middle of January. Sorry I wasn't able to respond sooner. The good news though is that, meanwhile, today, we were able to get approval from VT procurement to purchase news article data from the second source. We are going to start downloading real-time data from this source and will submit our request for the archived data (Oct 1, 2020 - Nov, 30 2020) ASAP. We hopefully will have that within the next two weeks. That will be processed and be able to be added to the current data (when ready) to hopefully improve model accuracy. I will send more details in the regularly scheduled update tomorrow but I wanted to get something back to you ASAP. Were you all able to determine if daily encounter data would be available? We asked about and found some here <https://www.cbp.gov/document/stats/us-border-patrol-nationwide-apprehension…> but only up until Q4 of 2021. Were you able to find out if new (more up-to-date) data would be released? You also shared an Executive Leadership Report that I believe had daily and weekly encounter data. The raw numbers driving those reports could be very beneficial. Thanks, Brian On Wed, Dec 7, 2022 at 4:07 PM Flynn A Rapavi <FLYNNAR(a)dni.gov> wrote: Brian, Hope you are doing well. Just following up regarding the news article/data procurement agreement – were you able to finalize and begin data ingest? As you can imagine, we are getting quite a few questions regarding data/information availability. Consequently, if you can provide a solid target date for initiating model runs, that would be ideal. Sincere thanks in advance for any information you can provide – and if you think I can help in any way, please reach out. Best regards, Alexander Rapavi PMM/TOC/CD NIM-WHT 703.275.2711 -- Brian Mayer bmayer(a)cs.vt.edu 540-231-5907 Sanghani Center for Artificial Intelligence and Data Analytics <https://sanghani.cs.vt.edu/> - Virginia Tech <https://vt.edu/> -- Brian Mayer bmayer(a)cs.vt.edu 540-231-5907 Sanghani Center for Artificial Intelligence and Data Analytics <https://sanghani.cs.vt.edu/> - Virginia Tech <https://vt.edu/> -- Brian Mayer bmayer(a)cs.vt.edu 540-231-5907 Sanghani Center for Artificial Intelligence and Data Analytics <https://sanghani.cs.vt.edu/> - Virginia Tech <https://vt.edu/>

1 1

Migration Notes from ITA
by Brian Mayer 12 Dec '22

12 Dec '22

- Repeat encounters vs unique encounters are available in their DHS DB but is CUI and requires DHS computer - It sounds like they know how many enter the US but that might be in the CUI DB. - They are using Socio-political data (e.g., OECPs) - They are using moving monthly averages as a feature because trends tend to be significantly longer than 1 month - They can't pick up the base causes for bumps (we should have better luck using news as a data source and causal analysis) - The DHS metric of interest was not known but they eventually settled on MAPE, which was similar to RMSE/actual value

1 0

Fwd: Status Request
by Brian Mayer 12 Dec '22

12 Dec '22

---------- Forwarded message --------- From: Flynn A Rapavi <FLYNNAR(a)dni.gov> Date: Mon, Dec 12, 2022 at 11:33 AM Subject: RE: Status Request To: Brian Mayer <bmayer(a)cs.vt.edu>, Naren Ramakrishnan <naren(a)cs.vt.edu> Cc: Nicole C Peruyera <NICOLECP(a)nctc.gov>, Flynn A Rapavi <FLYNNAR(a)dni.gov> Brian, Definitely appreciate the specifics. Data volume is impressive and will likely yield some beneficial insights- decidedly excited about the output. As you can imagine, with pending changes to specific provisions in US immigration law (e.g. starting 21 Dec timeframe), we’ve been getting some questions (from internal, external sources) regarding possible projections in the lead-up. Do you have the ability to execute some prelim runs based on existing data set availability (e.g. prior to 21 Dec)? Would said runs (even if limited in duration) be executable this week? Recognize the limitations in doing this, but would be beneficial for us to inform conversations and for comparative purposes. Regarding your questions, I’ve reached out to my contacts twice and have not heard back. I will call them this week and see if I can lock down whether data post 4Q21 can be made available and also if they can do a direct push of the executive leadership reports- agree, getting the raw data would be ideal. Will see what I can do— Thanks for the big forward-lean on this project— V/R, Alexander R. PMM/TOC/CD DMI/NIM-WHT 703.275.2711 *From:* Brian Mayer <bmayer(a)cs.vt.edu> *Sent:* Thursday, December 08, 2022 8:06 PM *To:* Flynn A Rapavi <FLYNNAR(a)dni.gov>; Naren Ramakrishnan <naren(a)cs.vt.edu> *Cc:* Nicole C Peruyera <NICOLECP(a)nctc.gov> *Subject:* Re: Status Request Alex, We have already ingested 2.8M articles from Oct 1, 2020 - current from the one source and we are ingesting 3000+ articles per day coming from over 4000 potential domains. We are finishing up the enrichment of these articles to prepare them to be used by the forecasting model. This involves extracting entities and creating keyword counts. We should be ready to start giving forecasts by the middle of January. Sorry I wasn't able to respond sooner. The good news though is that, meanwhile, today, we were able to get approval from VT procurement to purchase news article data from the second source. We are going to start downloading real-time data from this source and will submit our request for the archived data (Oct 1, 2020 - Nov, 30 2020) ASAP. We hopefully will have that within the next two weeks. That will be processed and be able to be added to the current data (when ready) to hopefully improve model accuracy. I will send more details in the regularly scheduled update tomorrow but I wanted to get something back to you ASAP. Were you all able to determine if daily encounter data would be available? We asked about and found some here <https://www.cbp.gov/document/stats/us-border-patrol-nationwide-apprehension…> but only up until Q4 of 2021. Were you able to find out if new (more up-to-date) data would be released? You also shared an Executive Leadership Report that I believe had daily and weekly encounter data. The raw numbers driving those reports could be very beneficial. Thanks, Brian On Wed, Dec 7, 2022 at 4:07 PM Flynn A Rapavi <FLYNNAR(a)dni.gov> wrote: Brian, Hope you are doing well. Just following up regarding the news article/data procurement agreement – were you able to finalize and begin data ingest? As you can imagine, we are getting quite a few questions regarding data/information availability. Consequently, if you can provide a solid target date for initiating model runs, that would be ideal. Sincere thanks in advance for any information you can provide – and if you think I can help in any way, please reach out. Best regards, Alexander Rapavi PMM/TOC/CD NIM-WHT 703.275.2711 -- Brian Mayer bmayer(a)cs.vt.edu 540-231-5907 Sanghani Center for Artificial Intelligence and Data Analytics <https://sanghani.cs.vt.edu/> - Virginia Tech <https://vt.edu/> -- Brian Mayer bmayer(a)cs.vt.edu 540-231-5907 Sanghani Center for Artificial Intelligence and Data Analytics <https://sanghani.cs.vt.edu/> - Virginia Tech <https://vt.edu/>

1 0

VT Migration Prediction Update
by Brian Mayer 09 Dec '22

09 Dec '22

All, Please find attached our biweekly status report. Let us know if there is anything you would like to discuss further. Thanks, Brian (and team) -- Brian Mayer bmayer(a)cs.vt.edu 540-231-5907 Sanghani Center for Artificial Intelligence and Data Analytics <https://sanghani.cs.vt.edu/> - Virginia Tech <https://vt.edu/>

1 0