Hivemind: Persistent Trends

Hivemind persistent trends, or trends aggregated by similarity across time, are designed to capture narratives that remain relevant over time.

They reduce short-term noise and emphasize trends that exhibit consistency, stability, and persistence across market cycles.

UI Access

Persistent trends are available in the ForecastOS UI, where users can compare persistent and raw trends, visualize long-run narratives, and evaluate trend durability through time.

Hivemind Persistent Trends

API Access

Hivemind persistent trends are accessible through the ForecastOS API for systematic research and portfolio workflows.

curl -X GET "https://app.forecastos.com/api/v1/persistent_trends" \
-H "Authorization: Bearer YOUR_API_TOKEN" \
-H "Content-Type: application/json"

Query Parameters

Parameter Type Default Required Description
page integer 1 No Page number for paginated results.
min_days_market_relevant integer 1 No Minimum number of days a trend must be flagged as market relevant.
filter_start_date string 2 years ago (UTC, date) No Start date (inclusive) for filtering trends by date.
filter_end_date string Today (UTC, date) No End date (inclusive) for filtering trends by date.

Response

{
  "data": [
    {
      "id": 1689,
      "title": "ai, artificial, generative, ais",
      "last_seen_at": "2025-11-10T22:22:01.039Z",
      "instance_first_seen": "2024-02-01T00:00:00.000Z",
      "instance_last_seen": "2025-11-10T00:00:00.000Z",
      "weighted_instance_count": 649
    }
  ],
  "meta": {
    "page": 1,
    "per_page": 100,
    "total_count": 118,
    "total_pages": 2
  }
}

Open-Source Access

The open-source ForecastOS Python library includes utilities to retrieve persistent trend data and apply common transformations.

Persistent trend construction logic and source weighting are handled within the managed Hivemind platform.

import forecastos as fos

df_persistent_trend = fos.PersistentTrend.get_df(
  min_days_market_relevant=100,
  filter_start_date='2020-01-01',
  filter_end_date='2025-01-01'
)

Parameters

Parameter Type Default Required Description
min_days_market_relevant integer 1 No Minimum number of days a trend must be flagged as market relevant.
filter_start_date string 2 years ago (UTC, date) No Start date (inclusive) for filtering trends by date.
filter_end_date string Today (UTC, date) No End date (inclusive) for filtering trends by date.

This returns a time-series DataFrame for all the persistent trends matching the filter parameters.

Let's explore Hivemind Custom Trends next.