What is No Code?

The Taiyō Toolkit functions as the engine providing standards for domain-specific analysis and building prediction pipelines. With the Taiyō Toolkit, customers can leverage

  • User-friendly time series AI algorithms to forecast trends, identify drivers for daily, weekly, monthly, and monitor target indicators.
  • The toolkit’s fundamental value comes as a time and cost-effective solution to the research and due diligence required to dynamically monitor global trends.

How Does it Works?


Data curation is a painful and cumbersome analytical process. The data process and pipeline include[s] the following steps:

Data Ingestion

  • Write scripts to clean and ingest the data.
  • Have a system in place to dynamically update all the data on a given frequency
  • Clean, preprocess, merge the different data
  • Generate new features from existing data

Analysis

  • Visualize feature space and conduct exploratory data analysis (EDA)
  • Select relevant features with unsupervised learning and clustering
    • Similarity, clustering, pattern causality, correlations to help identify dynamic factors driving the target
  • Benchmark and evaluate the top prediction model: Based on various regression or classification performance metrics, evaluate which model to trust more.

Prediction models

  • Write various models from scratch, including complex deep neural network architecture or classical models Ensemble the top models
  • Quantify uncertainty: Time-series forward forecasts should have confidence bounds associated with them and some indication of what the model does not know
  • Autotune the parameters: Models need to be fine-tuned based on the feature set and dynamically adjusting as the target changes, for example, predicting US monthly imports versus Germany exports models

Visualization

  • Production: Push the saved analysis production bucket so the data and model update on a given frequency autonomously, for example, daily or weekly basis
  • Explore more custom models, for example, low-frequency data like quarterly or annual can be appended based on daily movements, or build now-casting for various countries or trade outcomes significantly faster than current research practices.

Taiyō’s no-code and low-code interface provide actionable insights at your fingertips, bypassing the above complexities.

Highly flexible, the Taiyō Toolkit can be run on both a local computer or cloud infrastructure. Domain and business experts, economists, and data scientists can auto-generate dashboard reports with uncertainty bounded predictions for specified time series indicators. This enables domain experts, analysts, and data scientists to leverage global data to generate an asset-specific prediction, driver, and monitoring insights 90% faster. Similarly, alert features can be configured to obtain a summary of the main factors driving outcomes of interest.

The full life cycle of analysis becomes streamlined with an engine that provides custom analytics based on curated global data.