LISSA - Layer of Integration with Scikit-learn and Signal Analysis

LISSA is a Python library for standardized, reproducible scientific plotting, with a strong focus on time series visualization, comparative analysis, and declarative configuration via JSON.

Its main goal is to separate visualization logic from analysis code, allowing plots to be configured, reused, and audited without modifying plotting scripts.

A secondary goal is to provide examples of how to apply PCA methods, HMM methods and doing a general data analysis of the Equinor provided data of the main project. You can check it in examples folder.

---

Motivation

In data analysis and signal processing projects, it is common that:

• plotting code is tightly coupled to analysis logic;
• visual parameters are duplicated across scripts;
• reproducibility depends on undocumented global matplotlib state;
• style changes require touching multiple files.

LISSA addresses these issues by:

• encapsulating figure state in a single class;
• centralizing plot configuration in JSON files;
• using a fluent (chainable) API;
• leveraging matplotlib.rcParams in a controlled way.

---

Key Features

• Declarative configuration via JSON
• Fluent interface (.method().method().method())
• Native support for time series
• Direct integration with pandas and matplotlib
• Optional label translation and unit handling
• Centralized control via rcParams
• Clear separation between:
  • configuration
  • figure creation
  • plotting methods
  • finalization/export

---

Project Structure

.
├── src/
│   ├── lissa/
│       ├── __init__.py
│       ├── ml.py
│       ├── pca.py
│       ├── picture.py
│       ├── picture_classes.py
│       ├── processing.py

src/lissa/

Core library of the project. Contains all reusable code: including:

• signal processing utilities

• PCA and statistical analysis

• machine learning helpers

• plotting and figure abstraction (LissaFigure)

#├── examples/

examples/ Notebooks and scripts demonstrating how to use the library. Includes:

• PCA analysis

• HMM and GMM experiments

• filtering and scaling examples

• deprecated exploratory notebooks kept for reference

Subfolders:

method_examples/: focused examples of individual methods

plots/: JSON configuration files used by LissaFigure

dictionaries/: translation dictionaries and metadata

dynamax/: experiments using the dynamax library

├── docs/

Auto-generated documentation (HTML). typically produced by tools such as Sphinx or pdoc. Includes API documentation for the lissa modules.

├── pyproject.toml

Project configuration file defining:

• dependencies

• build system

• package metadata

├── README.md

Main project documentation.

---

Core Concept: JSON-Driven Configuration

Most visual behavior is defined in an external JSON file:

{
  "rc_params": {
    "figure.figsize": [18, 10],
    "font.size": 12,
    "axes.titlesize": 16,
    "legend.fontsize": 14
  },
  "plot_title": "Time Series Example",
  "log_scale_y": false
}

This file is loaded directly by the LISSA class and applied consistently.

---

Basic Usage Example

import lissafigure as li

fig = (
    li.LissaFigure("./plots/time_series_plot.json")
      .set_translation()
      .set_figure()
      .time_series_plot(data, index=0)
      .set_axes_texts(0, "Signal", "Time", "bar")
      .finalize()
)

---

Fluent Interface

All major methods return self, enabling chained calls:

(fig
 .set_figure()
 .time_series_plot(raw_data, 0)
 .time_series_plot(filtered_data, 1)
 .finalize())

---

Main Methods

set_figure()

• Creates the figure and axes

• Applies rcParams

• Sets layout and DPI

time_series_plot(data, index, ...)

• Plots time series on a selected axis

• Supports logarithmic scaling

• Uses pandas plotting backend

set_axes_texts(ax_or_idx, ...)

• Sets title, labels, and units

• Accepts either an Axes object or an axis index

finalize()

• Applies tight_layout

• Sets global title

• Saves the figure (if enabled)

---

Translation and Units

If configured, LissaFigure can:

• translate column names automatically;

• attach physical units to axis labels;

• switch between languages without modifying code.

---

Design Philosophy

• Configuration over code

• Explicit state control

• Reproducible graphics

• Minimal plotting boilerplate

• Low coupling with analysis logic

Dependencies

• Python ≥ 3.9

• matplotlib

• pandas

• numpy

---

Project Status

• Functional and actively used

• Ongoing simplification

• Progressive migration toward rcParams

• Public API still evolving

Roadmap

• Automatic API documentation

• Extended examples

• JSON schema validation

• Additional plot types