Machine Learning

未分类

Train/Test Split with Stratification Field Guide: When Five Input Decisions Decide Whether Your 95 Percent Accuracy Is Real or a 70/15/15 Illusion

By xiezixing 2026年9月30日

Stratification is what stops your 95 percent accuracy from being a 35 percent class misread.…

未分类

Train/Test Split with Stratification: A Field Guide When Class Balance Decides the Holdout

By xiezixing 2026年9月9日

Stratified train/test splitting is the difference between a model that reports 92% accuracy and one…

未分类

Min-Max scaling that respects the data you actually have

By xiezixing 2026年8月1日

The closing claim Most "Min-Max normalizer" tutorials skip the seam where things go wrong: missing…

未分类

Z-Score Standardization, Done Without a Spreadsheet

By xiezixing 2026年7月29日

Z-Score Standardization, Done Without a Spreadsheet A column of salaries and a column of ages…

未分类

Why Your Correlation Matrix Lies About Non-Linear Data: The 3-Statistic Trap

By xiezixing 2026年7月21日

What if the spreadsheet that tells you two stocks "strongly correlate" is actually measuring a…

Regression Analyzer — R-Squared Will Lie to You cover image

未分类

R-Squared Will Lie to You: A Field Guide to Reading Regression Output

By xiezixing 2026年7月10日

Five numbers matter more than R-squared — and in this order. Why held-out validation, residual plots, and prediction intervals beat the headline score every time, and a practical workflow for shipping models that survive production.

未分类

The five boundary decisions that quietly break every model

By xiezixing 2026年6月28日

Every data pipeline eventually meets a value it does not expect. The temperature sensor reports…

Min-Max Normalizer poster: gradient bar from min=0 to max=1, DATA PROCESSING tag

未分类

The Three Things Min-Max Scaling Quietly Does in Your ML Pipeline

By xiezixing 2026年6月25日

If you have ever trained a model and watched one feature dominate the loss curve…

未分类

Dataset Imbalance: The Three Ratios Your Spreadsheet Liked Until Your Model Did Not

By xiezixing 2026年6月22日

Three numbers to compute before training a classifier, five strategies that actually fix class imbalance, and the four cases when you should not resample at all.

未分类

The 73 Score That Quietly Decides Whether Your Dataset Is Production-Ready

By xiezixing 2026年6月20日

The 73 Score That Quietly Decides Whether Your Dataset Is Production-Ready A dataset profile is…