I need to find articles in which the authors trained machine learning model(s) with data from one geographical place, and, afterwards, applied the trained model on data about another geographical place. The need to do that is to know if the model can generalize well with unseen data, like, missing data (but I don't have any data missing).
I've tried asking suggestions for: AnswerThis, Perplexity, DeepSeek, ChatGPT, and Gemini. However, the AIs' suggested keywords did not lead me to the gold mine I was expecting to find.
Which keywords would best lead me to the articles I'm looking for?
Very relevant articles found (and read) after opening this question:
- (Zhou, 2023) — "Domain Generalization: A Survey";(Deng, 2024) — "Domain Generalization in Time Series Forecasting"
They, however, are still not specific enough to solve my problem: they are not about the source domain being a time series of one city, and the target domain being the time series of another city.
Interesting keywords taken from TITLE-ABS-KEY of the very relevant articles above:
- Out-of-Distribution Generalization;Domain Shift;Model Robustness;Domain Generalization;OOD Generalization;Time Series Forecasting;Time Series Analysis;Regularization;Unseen Target Domains;Observed Source Domains;Time Series Domains;Distributions Shifts;Domain Discrepancy Regulatization.
Search strings I've tried so far:
The number before the string is the number of articles returned on SCOPUS.
------------------------------------------------------------------------TITLE-ABS-KEY ( A ) AND NOT TITLE-ABS-KEY ( B ):------------------------------------------------------------------------313 A = "Domain Generaliz*" AND ( "Drought" OR "SPEI" OR "SPI" OR "Forecast*" OR "Regression" OR "Time Series" OR "Timeseries" OR "Weather" OR "Meteorolog*" OR "Precipitation" ) B = ""NOTE: missing ( "Predict*" OR "Environment*" )------------------------------------------------------------------------143 A = "Cross-Domain Generaliz*" AND ( "Drought" OR "SPEI" OR "SPI" OR "Forecast*" OR "Regression" OR "Time Series" OR "Timeseries" OR "Weather" OR "Meteorolog*" OR "Precipitation" OR "Predict*" OR "Environment*" ) B = ""------------------------------------------------------------------------67 A = "Cross-Domain" AND ( "Drought" OR "SPEI" OR "SPI" OR "Timeseries" OR "Meteorolog*" OR "Precipitation") B = ""------------------------------------------------------------------------5 A = "Spa* Domain Generaliz*" OR "Geo* Domain Generaliz*" B = ""------------------------------------------------------------------------1 A = "Spa* Cross-Domain Generaliz*" OR "Geo* Cross-Domain Generaliz*" B = ""------------------------------------------------------------------------Keywords suggested by my particular AI's Committee:
- geographical generalization;transfer learning;cross-region generalization;cross-city transferability and generalization;explicitly testing transferability across cities;model transfer between cities;cross-city machine learning;formal transfer learning and domain adaptation approach;knowledge transfer;cross-city knowledge transfer;transfer across cities;cross-city transfer learning;pooling data across cities;spatial generalization;transfer from epicenter cities to other urban neighborhoods;domain adaptation;spatial correlation;domain adaptation between cities;spatial transfer learning;geographical generalization;cross-location generalization;spatial transferability;geographic transferability;spatial generalization;out-of-region prediction;cross-region model transfer;spatial domain adaptation;spatial domain generalization;geographic transfer learning;geographic domain adaptation;spatial generalization;geographic domain adaptation;geographic transfer learning.
