MLPicks

By rachanapaudel26

0+ active installationsVersion 1.0.0

Smart related posts and tailored content recommendations powered by a lightweight ML engine. Zero cloud APIs, instant loading, and privacy-first.

Description

Are you struggling to keep readers on your WordPress site after they finish an article?

Most visitors read a single blog post and leave right away. If your website is still relying on basic category lists or slow plugins that show random stories, you are leaking visitors, page views, and revenue every single day.

If you aren’t showing your audience what to read next with fast and relevant suggestions, your hard-earned traffic is walking straight out the door.

Keep Readers on Your Site with MLPicks

Keep your readers engaged and dramatically increase on-site user interaction with intelligent, personalized content recommendations powered by a lightweight machine learning engine.

By combining ML power with quiet background processing, MLPicks shows your visitors articles they actually want to read next. Instead of slowing down your website with heavy database searches, it prepares recommendations in advance so your pages load instantly.

Everything runs 100% on your own server in pure PHP. There are no monthly fees, no external accounts to set up, and zero tracking cookies.

The Problem: Why Ordinary Related Posts Plugins Fail

Most site owners try to keep visitors around using standard related content plugins, but run into the same frustrating issues:

  1. They Slow Down Your Site: Traditional plugins search your database every time a visitor opens a page. As your site gets more articles and traffic, your pages load slower and your site speed scores drop.
  2. They Show Unrelated Posts: Most plugins simply grab the newest 3 posts from the same category. A reader finishing a specific guide often gets shown totally unrelated content simply because both share a general tag.
  3. They Never Learn: Standard plugins show the same static lists to everyone. They cannot adapt to what your readers are actually interested in.
  4. Third-Party AI Tools Are Costly and Heavy: External AI widgets require monthly subscriptions, inject heavy tracking scripts into your site, and send your visitor data to third-party servers.

The Solution: Smart Recommendations That Never Slow You Down

MLPicks is built differently. It does all the heavy thinking quietly in the background so your frontend stays lightning fast:

  • Instant Page Loads: Recommendations are pre-calculated in the background. When someone visits a post, suggestions load in under a millisecond.
  • Smart Content Matching: Evaluates actual article content, reader journeys, and freshness—not just broad categories.
  • On-Device Reader Tailoring: Adapts to what each reader has been exploring during their visit, keeping all data private inside their own browser.
  • Total Data Privacy: No tracking cookies, no IP logging, and no visitor data ever leaves your server.

Machine Learning Architecture: How the ML Engine Works

Instead of locking up your MySQL database with slow queries, MLPicks combines mathematical vector space modeling, reader behavior paths, and logistic regression to deliver genuinely relevant suggestions at sub-millisecond speeds.

MLPicks is not a superficial wrapper around an external API. It features a genuine, self-hosted machine learning engine implemented in native PHP:

1. TF-IDF Vector Space & Semantic Term Rarity
MLPicks tokenizes your titles, excerpts, and content into mathematical term frequency-inverse document frequency (TF-IDF) vectors. It incorporates BM25-style rarity weighting so unique, informative topics carry higher mathematical weight than generic filler words.

2. 5-Feature Logistic Regression Scoring
Candidate posts are scored through a multivariate logistic regression model with the probability function $P(\text{click}) = \sigma(w_0 + \sum w_i x_i)$:
* $x_1$ – Content Similarity: Cosine similarity between candidate post vectors and the active article or reader interest centroid.
* $x_2$ – Markov Transition Probability: First-order Markov chain transitions calculated from aggregate, anonymous reader click paths ($P(\text{Post}B \mid \text{Post}_A)$).
* $x_3$ – Popularity Velocity: Log-normalized 30-day view velocity ($\ln(1 + \text{views}
{30\text{d}}) / \ln(1 + \text{max_views}_{30\text{d}})$), surfacing trending content without letting older posts monopolize recommendations.
* $x_4$ – Taxonomy Affinity: Jaccard overlap between taxonomy terms, weighted by inverse category frequency.
* $x_5$ – Freshness Decay: Continuous exponential half-life decay ($\exp(-\lambda \cdot \Delta t)$ with a 30-day half-life), ensuring newer posts receive a gentle, natural boost.

3. Continuous Online Learning via Mini-Batch SGD
The model automatically refines itself. As visitors interact with recommendations, an asynchronous background task runs Mini-Batch Stochastic Gradient Descent (SGD) with L2 regularization ($\lambda = 0.001$, $\eta = 0.05$). The weights ($w_0 \dots w_5$) continuously optimize for higher engagement without manual tuning.

4. Specialized Side-by-Side Comparison Model
MLPicks includes dedicated model weights calibrated specifically for comparison content (e.g., “Product A vs Product B”), identifying natural alternative choices and presenting them in high-converting comparison cards.

Key Features

  • Precomputed Background Matching: All similarity scoring, vector generation, and matrix calculations execute quietly via WordPress background cron jobs. Your frontend remains blazing fast.
  • Dynamic Grid & List Layouts: Present suggestions in clean, modern card grids or sleek horizontal list rows that seamlessly match any theme’s typography and color scheme.
  • Client-Side Dynamic Discovery: With the [mlpicks_recommendations] shortcode, readers receive individualized suggestions calculated from their recent browsing session, without using server-side tracking cookies.
  • Side-by-Side Comparison (VS) Cards: Automatically link product reviews and comparative articles with the [mlpicks_comparison] shortcode, complete with thumbnail teasers and feature highlights.
  • Intelligent Thumbnail Fallbacks: Automatically detects in-content images if no featured image is set, or falls back to an elegant typography-focused card layout to prevent broken images.
  • Editorial Control & Hand-Picked Overrides: Pin specific high-priority articles directly from the post editor sidebar, or hide recommendation boxes on specific landing pages with one click.
  • Built-in Analytics & Suggestion Health: Monitor impressions, clicks, click-through rates (CTR), and identify unindexed or orphaned articles from the intuitive admin report.
  • 100% Cache & CDN Compatible: Compatible with WP Rocket, LiteSpeed Cache, W3 Total Cache, Fastly, Cloudflare, and server-level object caches (Redis, Memcached).
  • Multisite & Multi-Post-Type Ready: Works seamlessly with custom post types and WordPress Multisite networks.

How It Works

  1. Indexing: When activated, MLPicks catalogs your published articles in the background, tokenizing content and building the TF-IDF term matrix.
  2. Matching: The 5-feature ML algorithm calculates correlation scores between articles and stores top candidates in an indexed database table (wp_mlpicks_related).
  3. Instant Display: When a reader views an article, MLPicks fetches precomputed matches with an indexed, single-row query (sub-millisecond execution).
  4. Learning: As visitors view and click recommended articles, anonymous interaction counts update aggregate transition probabilities. The SGD optimizer periodically tunes model weights for superior recommendation accuracy.

Use Cases

  • Content Creators & Bloggers: Keep readers engaged on your site longer, reducing bounce rates and boosting ad impressions.
  • Affiliate & Review Sites: Connect related product reviews and comparison guides automatically, driving higher buyer intent and affiliate conversions.
  • News & Digital Magazines: Deliver trending stories and topical follow-ups without straining your web server during breaking traffic spikes.
  • Corporate & Niche Documentation: Guide users effortlessly to relevant documentation, whitepapers, and troubleshooting articles.

Why Choose MLPicks Over Conventional Alternatives?

  • vs. Traditional Related Post Plugins: Standard plugins execute real-time SQL queries that slow down page loads. MLPicks precomputes matches in the background for zero frontend impact.
  • vs. Cloud-Based SaaS Recommendations: Third-party AI tools charge expensive monthly subscription fees and send your visitor data offsite. MLPicks is 100% free, self-hosted, and privacy-first.
  • vs. Generic Category Lists: Category-based widgets show chronological posts regardless of relevance. MLPicks uses multi-factor machine learning to surface articles readers genuinely want to read.

Getting Started

  1. Activate the plugin.
  2. Background indexing starts automatically—no complex setup required.
  3. Recommendations immediately appear below your posts.
  4. Customize your layout, card count, and headings under MLPicks > Settings.

Shortcodes

MLPicks includes a versatile set of shortcodes to place recommendations anywhere in your content, templates, or block layouts:

[mlpicks]

Renders algorithmic related articles based on the current post.
* count (int): Number of cards to display (default: 4, range: 1–12).
* heading (string): Custom section title (e.g., heading="Further Reading").
* layout (string): Card presentation style: grid (default) or list.
* show_date (string): Display publication dates: yes or no (default: no).
* fallback (string): Fallback behavior when few matches exist: hide or latest.

Example: [mlpicks count="3" layout="grid" heading="Recommended Stories"]

[mlpicks_recommendations]

Displays dynamic content recommendations curated to the individual visitor’s reading interests.
* count (int): Number of cards to display (default: 4).
* heading (string): Custom section title (default: Recommended for you).
* layout (string): grid or list.

Example: [mlpicks_recommendations count="4" layout="grid"]

[mlpicks_most_read]

Displays your site’s most popular articles based on aggregate interaction metrics.
* count (int): Number of articles to display (default: 4).
* period (int|string): Time window in days: 7, 30, or all (default: 30).
* heading (string): Custom title (default: Trending Now).
* layout (string): grid or list.

Example: [mlpicks_most_read count="4" period="7" heading="This Week's Top Stories"]

[mlpicks_recent_articles]

Displays the newest published articles from your archive.
* count (int): Number of articles to display (default: 4).
* heading (string): Custom section title (default: Recent articles).
* layout (string): grid or list.

Example: [mlpicks_recent_articles count="3" layout="list"]

[mlpicks_comparison]

Displays specialized side-by-side comparison cards for product reviews and comparative articles.
* count (int): Number of comparison cards to display (default: 2).
* heading (string): Custom section title (default: Compare Alternatives).
* layout (string): grid or list.

Example: [mlpicks_comparison count="2" layout="grid"]

Installation

Automatic Installation

  1. Log in to your WordPress administrator dashboard.
  2. Navigate to Plugins > Add New.
  3. Search for MLPicks.
  4. Click Install Now, then click Activate.

Manual Installation

  1. Download the mlpicks.zip archive.
  2. Extract the folder and upload mlpicks to the /wp-content/plugins/ directory of your WordPress installation.
  3. Activate the plugin through the Plugins menu in WordPress.

After Activation

  • MLPicks immediately begins building the semantic vector index in the background using WordPress cron.
  • To customize appearance, navigate to MLPicks > Settings in your WordPress admin menu.
  • To view performance metrics, visit MLPicks > Report.

FAQ

Does MLPicks slow down my website or increase server load?

No. MLPicks is engineered specifically to eliminate the performance penalties common to related posts plugins. All mathematical modeling, vector comparisons, and learning routines run asynchronously during idle background cron intervals. Frontend display relies on an indexed database lookup that executes in under 1 millisecond.

How does MLPicks protect visitor privacy and comply with GDPR / CCPA?

MLPicks does not store IP addresses, does not use tracking cookies, does not collect personally identifiable information (PII), and never transmits data to third-party cloud servers. Tailored reader suggestions rely on anonymous client-side vector weights stored in the visitor’s own browser storage. It is fully GDPR, CCPA, and ePrivacy compliant out of the box.

Does MLPicks work with page caching and CDNs?

Yes. Standard related post boxes render clean, static HTML that is 100% compatible with page caching plugins (such as WP Rocket, LiteSpeed Cache, W3 Total Cache) and CDN edge caching (Cloudflare, Fastly). When using the client-side tailored recommendations shortcode ([mlpicks_recommendations]), static placeholder cards are populated via lightweight JavaScript without busting your page cache.

What makes MLPicks different from plugins that just use post categories?

Category-based plugins perform simplistic queries that often display completely irrelevant articles that happen to share a broad taxonomy term. MLPicks evaluates content using TF-IDF term rarity, Markov visitor navigation probabilities, taxonomy Jaccard affinity, popularity velocity, and exponential freshness decay, combined through a trained logistic regression model.

How does the plugin handle new sites or articles with little traffic (Cold Start)?

For brand-new articles or sites with little traffic history, the algorithm leans heavily on textual TF-IDF cosine similarity, term rarity, and taxonomy affinity. As readers begin browsing, Markov transition probabilities and popularity velocity naturally calibrate to incorporate real user behavior.

Can I manually choose which posts appear on specific articles?

Yes. On any post edit screen, use the MLPicks – Hand-Picked Suggestions metabox to search and pin specific articles. Hand-picked selections always take top priority, with any remaining card slots automatically filled by algorithmic recommendations.

What happens if an article has no related content?

In MLPicks > Settings, you can choose between two fallback behaviors:
* Hide Box (Recommended): If no articles pass the relevance threshold, the recommendation container is cleanly omitted so your page never displays off-topic filler.
* Latest Posts: Automatically fills remaining card slots with your most recent published articles.

Can I exclude specific posts or categories?

Yes. You can globally exclude specific categories or individual post IDs under MLPicks > Settings. You can also check “Hide the Read next box on this post” directly within the post editor.

Does MLPicks require an OpenAI API key or external cloud subscription?

No. MLPicks runs entirely on your own WordPress server using optimized PHP algorithms. It requires no API keys, no external subscriptions, and incurs zero monthly operational costs.

Changelog

1.0.0

  • Initial public release.
  • Lightweight background matching engine for instant front-end page loads.
  • Smart topic matching combining categories, tags, text similarity, and freshness decay.
  • Local tailor-made recommendation engine based on reader interests.
  • Side-by-side comparison (VS) cards with automatic competitor matching.
  • Full shortcode suite: [mlpicks], [mlpicks_recommendations], [mlpicks_most_read], [mlpicks_recent_articles], and [mlpicks_comparison].
  • 100% privacy-safe daily click, view, and viewport impression tracking.
  • Comprehensive Click & Suggestion Report with helpful explanatory tooltips.
  • Hand-picked suggestion overrides and per-post visibility controls in the editor.
  • Multisite compatibility and clean uninstallation cleanup.