BigQuery ML を使用した ML モデルの作成: チャレンジラボ のレビュー

Marri Mahesh · 1日前にレビュー済み

SIngh anugrah · 1日前にレビュー済み

Miroshnyk Sergii · 2日前にレビュー済み

Branco Xavier Clauderson · 2日前にレビュー済み

sameeha zumlath · 2日前にレビュー済み

RAMANI NIKHIL · 2日前にレビュー済み

GoogleUser1 Abhishek · 2日前にレビュー済み

Goswami2 Arijit · 2日前にレビュー済み

srirama manaswin · 2日前にレビュー済み

good

Nugroho Zaki Bimo · 2日前にレビュー済み

Juwita Abshari Jeni · 2日前にレビュー済み

Setp to evaluate an improved model is failing always

Marri Mahesh · 2日前にレビュー済み

Bari Sushant · 2日前にレビュー済み

thanks

Ardiyansa Bagus · 2日前にレビュー済み

Reddy Kiran · 2日前にレビュー済み

good lab

Ojha Jit · 2日前にレビュー済み

Enaytulla Sheikh · 2日前にレビュー済み

Marri Mahesh · 2日前にレビュー済み

Prasad Rohit · 2日前にレビュー済み

Choiri.A Brayen · 2日前にレビュー済み

Chaware Sarwaarth · 2日前にレビュー済み

The lab's core BQML syntax (CREATE MODEL, ML.PREDICT) is straightforward, but the surrounding task design has significant friction points that go beyond testing ML engineering skills: Undocumented domain knowledge required. The lab assumes prior familiarity with the Google Analytics BigQuery export schema — e.g. mapping hits.eCommerceAction.action_type numeric codes (0-8) to their meaning, or knowing that "traffic source" in the prompt actually maps to channelGrouping rather than the more intuitively-named trafficSource column. None of this is documented in the lab itself; it requires external research or prior GA360 experience.Mismatch with stated learning objective. This is positioned as an "ML Engineer" cert path (model deployment focus), but a large share of the lab's difficulty is data-analyst-level work: schema archaeology, e-commerce funnel semantics, and column disambiguation — not ML deployment itself. BQML's train/predict duplication isn't called out as a limitation. Feature engineering logic must be manually duplicated between the CREATE MODEL query and the ML.PREDICT query, with no pipeline abstraction (unlike sklearn/Vertex AI pipelines). This isn't explained anywhere in the lab, so a mismatch (e.g. missing a feature column) only surfaces as an opaque runtime error. Suggestion: Either provide a schema reference / glossary as part of the lab material, or explicitly frame this lab as also testing data exploration skills — not just ML deployment.

LAVILLEY VINCENT · 2日前にレビュー済み

Salam Aldi · 2日前にレビュー済み

laelafi dinda · 2日前にレビュー済み

Raghav Abhishek · 2日前にレビュー済み