Spatiotemporal Forecasting At Scale
Bibliographic record
Abstract
Spatiotemporal forecasting can be described as predicting the future value of a variable given when and where it will happen. This type of forecasting task has the potential to aid many institutions and businesses in asking questions, such as how many people will visit a given hospital in the next hour. Answers to these questions have the potential to spur significant socioeconomic impact, providing privacy-friendly short-term forecasts about geolocated events, which in turn can help entities to plan and operate more efficiently. These seemingly simple questions, however, present complex challenges to forecasting systems. With more GPS-enabled devices connected every year, from smartphones to wearables to IoT devices, the volume of collected spatiotemporal data that accompanies these questions has exploded, following the Big Data trend. This thesis proposes a forecasting framework that employs distributed computing in order to scale its internal components and overcome this high data volume scenario. It also designs discretization components that allow for flexibility in the framing of the forecasting questions. Furthermore, it devises a Geographically Global Model (GGM) backed by an ensemble of Stochastic Gradient Boosted Trees, a collection of Geographically Local Models (GLMs) backed by ARIMA models, and a non-linear blending of those as part of its multistage machine learning pipeline in order to boost its performance and stability. The merit of the proposed research is evaluated in three experiments, each of which comprises millions of records, namely forecasting hourly taxi demand in the city of New York, forecasting daily crime density in the city of Chicago, and forecasting hourly visits to places of interest across Canada. The experimental results show the effectiveness of the proposed Spatiotemporal Forecasting Framework in forecasting stable results across the three domains, while also outperforming the naive baseline by at least 49.8% with respect to the SMAPE residuals.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.001 |
| Open science | 0.001 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.002 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".