Matplotlib-rendered PNG plot images generated from sliding windows over financial time series features. Each image sample includes 1–8 day trend labels derived from the CL=F_Close price at the last index of the plotted window. The dataset, authored by mrzdu, is stored as 210 sharded Parquet files on Hugging Face and was last updated on 2026-01-04.
Use Cases
- Training image-based models for financial trend classification based on the 1–8 day labels.
- Developing visual pattern recognition systems for technical analysis of oil price charts.
- Benchmarking multimodal models that combine time-series data with image representations.
- Exploring sliding window techniques for generating labeled image sequences from financial data.
Strengths
- Dataset structure is explicitly defined with image samples paired with derived trend labels.
- Data is stored in a modern, sharded Parquet format optimized for distributed processing.
- The description specifies a concrete labeling method based on the CL=F_Close price.
Limitations
- Column-level documentation is absent; field semantics must be inferred after download.
- Row count, file size, and license information are unknown, limiting suitability assessment.
- Freshness should be verified as the last update timestamp is in the future (2026-01-04).
Provenance
- Source
- Hugging Face dataset repository authored by mrzdu.
- Collection Method
- Images generated by applying sliding windows to financial time series features and rendering plots with Matplotlib.
- Time Range
- null
- Freshness
- Last updated 2026-01-04 11:49:38
- Geography
- null