{"id":103297,"date":"2026-07-28T07:20:33","date_gmt":"2026-07-28T07:20:33","guid":{"rendered":"https:\/\/zamstudios.com\/blogs\/the-math-of-trust-using-shap-and-lime-to-demystify-black-box-credit-scoring-models\/"},"modified":"2026-07-28T07:20:33","modified_gmt":"2026-07-28T07:20:33","slug":"the-math-of-trust-using-shap-and-lime-to-demystify-black-box-credit-scoring-models","status":"publish","type":"post","link":"https:\/\/zamstudios.com\/blogs\/the-math-of-trust-using-shap-and-lime-to-demystify-black-box-credit-scoring-models\/","title":{"rendered":"The Math of Trust: Using SHAP and LIME to Demystify Black-Box Credit Scoring Models."},"content":{"rendered":"<div id=\"ez-toc-container\" class=\"ez-toc-v2_0_85 ez-toc-wrap-left counter-hierarchy ez-toc-counter ez-toc-grey ez-toc-container-direction\">\n<div class=\"ez-toc-title-container\">\n<p class=\"ez-toc-title\" style=\"cursor:inherit\">Table of Contents<\/p>\n<span class=\"ez-toc-title-toggle\"><a href=\"#\" class=\"ez-toc-pull-right ez-toc-btn ez-toc-btn-xs ez-toc-btn-default ez-toc-toggle\" aria-label=\"Toggle Table of Content\"><span class=\"ez-toc-js-icon-con\"><span class=\"\"><span class=\"eztoc-hide\" style=\"display:none;\">Toggle<\/span><span class=\"ez-toc-icon-toggle-span\"><svg style=\"fill: #999;color:#999\" xmlns=\"http:\/\/www.w3.org\/2000\/svg\" class=\"list-377408\" width=\"20px\" height=\"20px\" viewBox=\"0 0 24 24\" fill=\"none\"><path d=\"M6 6H4v2h2V6zm14 0H8v2h12V6zM4 11h2v2H4v-2zm16 0H8v2h12v-2zM4 16h2v2H4v-2zm16 0H8v2h12v-2z\" fill=\"currentColor\"><\/path><\/svg><svg style=\"fill: #999;color:#999\" class=\"arrow-unsorted-368013\" xmlns=\"http:\/\/www.w3.org\/2000\/svg\" width=\"10px\" height=\"10px\" viewBox=\"0 0 24 24\" version=\"1.2\" baseProfile=\"tiny\"><path d=\"M18.2 9.3l-6.2-6.3-6.2 6.3c-.2.2-.3.4-.3.7s.1.5.3.7c.2.2.4.3.7.3h11c.3 0 .5-.1.7-.3.2-.2.3-.5.3-.7s-.1-.5-.3-.7zM5.8 14.7l6.2 6.3 6.2-6.3c.2-.2.3-.5.3-.7s-.1-.5-.3-.7c-.2-.2-.4-.3-.7-.3h-11c-.3 0-.5.1-.7.3-.2.2-.3.5-.3.7s.1.5.3.7z\"\/><\/svg><\/span><\/span><\/span><\/a><\/span><\/div>\n<nav><ul class='ez-toc-list ez-toc-list-level-1 ' ><li class='ez-toc-page-1 ez-toc-heading-level-2'><a class=\"ez-toc-link ez-toc-heading-1\" href=\"https:\/\/zamstudios.com\/blogs\/the-math-of-trust-using-shap-and-lime-to-demystify-black-box-credit-scoring-models\/#LIME_The_Illusion_of_Local_Linearity\" >LIME: The Illusion of Local Linearity<\/a><\/li><li class='ez-toc-page-1 ez-toc-heading-level-2'><a class=\"ez-toc-link ez-toc-heading-2\" href=\"https:\/\/zamstudios.com\/blogs\/the-math-of-trust-using-shap-and-lime-to-demystify-black-box-credit-scoring-models\/#SHAP_Consistent_Explanations_from_Game_Theory\" >SHAP: Consistent Explanations from Game Theory<\/a><\/li><li class='ez-toc-page-1 ez-toc-heading-level-2'><a class=\"ez-toc-link ez-toc-heading-3\" href=\"https:\/\/zamstudios.com\/blogs\/the-math-of-trust-using-shap-and-lime-to-demystify-black-box-credit-scoring-models\/#Technical_Comparison_Evaluating_XAI_Frameworks\" >Technical Comparison: Evaluating XAI Frameworks<\/a><\/li><li class='ez-toc-page-1 ez-toc-heading-level-2'><a class=\"ez-toc-link ez-toc-heading-4\" href=\"https:\/\/zamstudios.com\/blogs\/the-math-of-trust-using-shap-and-lime-to-demystify-black-box-credit-scoring-models\/#Bridging_the_Gap_Between_Advanced_Mathematics_and_Industry_Need\" >Bridging the Gap Between Advanced Mathematics and Industry Need<\/a><\/li><li class='ez-toc-page-1 ez-toc-heading-level-2'><a class=\"ez-toc-link ez-toc-heading-5\" href=\"https:\/\/zamstudios.com\/blogs\/the-math-of-trust-using-shap-and-lime-to-demystify-black-box-credit-scoring-models\/#The_Path_to_Algorithmic_Transparency\" >The Path to Algorithmic Transparency<\/a><\/li><\/ul><\/nav><\/div>\n<p data-path-to-node=\"0\">When an algorithm denies a credit card application, a line of credit, or a mortgage, it doesn&#8217;t just reject a transaction; it reshapes a human trajectory. For decades, retail banking relied on transparent, highly interpretable statistical models like logistic regression and scorecards. If a customer was denied a loan, a credit risk officer could look at the precise coefficients and confidently say, &#8220;Your debt-to-income ratio is too high, and your utilization rate tripped a threshold.&#8221;<\/p>\n<p data-path-to-node=\"1\">However, traditional scoring models struggle to keep up with today&#8217;s complex financial data landscape. To squeeze maximum predictive accuracy out of millions of alternative data points\u2014ranging from utility payment histories to e-commerce transaction footprints\u2014the fintech industry has shifted toward ensemble methods like XGBoost, LightGBM, and deep neural networks.<\/p>\n<p data-path-to-node=\"2\">These models are exceptionally good at predicting default risk. But they operate as absolute black boxes.<\/p>\n<p data-path-to-node=\"3\">A neural network doesn&#8217;t give you a clean, linear reason for its decision; it processes inputs through hundreds of non-linear intersections, delivering a final probability score without context. In a highly regulated ecosystem governed by fair lending mandates and consumer protection laws, a simple &#8220;because the algorithm said so&#8221; is legally untenable and ethically reckless. The industry faces a deep tension: accept lower accuracy to keep things transparent, or maximize accuracy and risk regulatory penalties.<\/p>\n<p data-path-to-node=\"4\">To bridge this divide, explainable AI (XAI) frameworks have introduced mathematical models designed to extract transparency from black-box systems. By leveraging <b data-path-to-node=\"4\" data-index-in-node=\"162\">LIME<\/b> and <b data-path-to-node=\"4\" data-index-in-node=\"171\">SHAP<\/b>, risk teams can maintain cutting-edge predictive power while ensuring every automated credit decision is auditable, fair, and legally compliant.<\/p>\n<h2 data-path-to-node=\"6\"><span class=\"ez-toc-section\" id=\"LIME_The_Illusion_of_Local_Linearity\"><\/span>LIME: The Illusion of Local Linearity<span class=\"ez-toc-section-end\"><\/span><\/h2>\n<p data-path-to-node=\"7\"><b data-path-to-node=\"7\" data-index-in-node=\"0\">Local Interpretable Model-agnostic Explanations (LIME)<\/b> approaches the black-box problem through a simple yet powerful insight: even if a global machine learning model is wildly non-linear and impossible to understand across an entire dataset, its behavior around one specific individual\u2019s data point can be approximated by a simple, transparent model.<\/p>\n<div class=\"code-block ng-tns-c2503008527-39 ng-animate-disabled ng-trigger ng-trigger-codeBlockRevealAnimation\">\n<div class=\"formatted-code-block-internal-container ng-tns-c2503008527-39\">\n<div class=\"animated-opacity ng-tns-c2503008527-39\">\n<pre class=\"ng-tns-c2503008527-39\"><code class=\"code-container formatted ng-tns-c2503008527-39 no-decoration-radius\" role=\"text\" data-test-id=\"code-content\">[Global Non-Linear Decision Boundary] \n             \u2502\n             \u25bc\n   [Isolate Single Applicant] \u2500\u2500&gt; [Perturb Attributes &amp; Generate Synthetic Neighbors]\n             \u2502\n             \u25bc\n   [Train Simple Linear Surrogate Model Locally] \u2500\u2500&gt; [Extract Feature Weights]\n<\/code><\/pre>\n<\/div>\n<\/div>\n<\/div>\n<p data-path-to-node=\"9\">To explain a single applicant&#8217;s credit score, LIME runs a localized experiment:<\/p>\n<ol start=\"1\" data-path-to-node=\"10\">\n<li>\n<p data-path-to-node=\"10,0,0\">It takes the applicant&#8217;s real profile (e.g., age, income, existing debt) and generates a synthetic cluster of slightly altered data points around them.<\/p>\n<\/li>\n<li>\n<p data-path-to-node=\"10,1,0\">It pushes these synthetic profiles through the complex black-box model to see how the credit denial probability shifts.<\/p>\n<\/li>\n<li>\n<p data-path-to-node=\"10,2,0\">It weights these new samples based on how close they are to the real applicant, then trains a simple, highly interpretable linear surrogate model (like a basic ridge regression) on this localized dataset.<\/p>\n<\/li>\n<\/ol>\n<p data-path-to-node=\"11\">Mathematically, LIME optimizes the following objective function:<\/p>\n<div data-path-to-node=\"12\">\n<div class=\"math-block\" data-math=\"xi(x) = argmin_{g in G} mathcal{L}(f, g, pi_x) + Omega(g)\">$$xi(x) = argmin_{g in G} mathcal{L}(f, g, pi_x) + Omega(g)$$<\/div>\n<\/div>\n<p data-path-to-node=\"13\">Where <span class=\"math-inline\" data-math=\"f\" data-index-in-node=\"6\">$f$<\/span> represents the complex black-box model, <span class=\"math-inline\" data-math=\"g\" data-index-in-node=\"48\">$g$<\/span> is the simple, human-readable surrogate model (such as a sparse linear regression), and <span class=\"math-inline\" data-math=\"pi_x\" data-index-in-node=\"138\">$pi_x$<\/span> defines the proximity measure that dictates how far the synthetic variations are allowed to drift from the real individual <span class=\"math-inline\" data-math=\"x\" data-index-in-node=\"267\">$x$<\/span>. The term <span class=\"math-inline\" data-math=\"mathcal{L}(f, g, pi_x)\" data-index-in-node=\"279\">$mathcal{L}(f, g, pi_x)$<\/span> measures the unfaithfulness of the surrogate model <span class=\"math-inline\" data-math=\"g\" data-index-in-node=\"355\">$g$<\/span> in approximating the true model <span class=\"math-inline\" data-math=\"f\" data-index-in-node=\"389\">$f$<\/span> within that local neighborhood, while <span class=\"math-inline\" data-math=\"Omega(g)\" data-index-in-node=\"429\">$Omega(g)$<\/span> acts as a regularization penalty that keeps the surrogate model simple enough for a human credit officer to understand.<\/p>\n<p data-path-to-node=\"14\">LIME\u2019s greatest strength is its sheer speed and model-agnostic nature; it doesn&#8217;t care if your underlying architecture is a massive neural network or a complex gradient-boosted tree. However, it suffers from a significant drawback: <b data-path-to-node=\"14\" data-index-in-node=\"232\">instability<\/b>. Because it relies on random local sampling, running LIME twice on the exact same applicant can occasionally yield slightly different feature weights, a variance that can complicate formal legal compliance audits.<\/p>\n<h2 data-path-to-node=\"16\"><span class=\"ez-toc-section\" id=\"SHAP_Consistent_Explanations_from_Game_Theory\"><\/span>SHAP: Consistent Explanations from Game Theory<span class=\"ez-toc-section-end\"><\/span><\/h2>\n<p data-path-to-node=\"17\">When financial institutions require absolute mathematical consistency and legal auditability, <b data-path-to-node=\"17\" data-index-in-node=\"94\">SHapley Additive exPlanations (SHAP)<\/b> serves as the gold standard. Instead of sampling data points locally, SHAP adapts a proven concept from cooperative game theory: <b data-path-to-node=\"17\" data-index-in-node=\"260\">Shapley values<\/b>.<\/p>\n<p data-path-to-node=\"18\">Imagine a group of cooperative players working together to win a cash prize in a tournament. The players contribute unequally to the final score. Game theory asks: how do you distribute the prize money among the players in a way that is mathematically fair, accounting for both their individual skills and their collaborative synergies?<\/p>\n<p data-path-to-node=\"19\">In credit scoring, the &#8220;tournament prize&#8221; is the difference between the applicant\u2019s calculated default probability and the baseline average default probability across the entire population. The &#8220;players&#8221; are the applicant&#8217;s specific financial features\u2014such as credit utilization, debt-to-income ratio, and payment history.<\/p>\n<p data-path-to-node=\"20\">The marginal contribution of feature <span class=\"math-inline\" data-math=\"i\" data-index-in-node=\"37\">$i$<\/span> is calculated across all possible subsets of features that can be formed from the total set <span class=\"math-inline\" data-math=\"N\" data-index-in-node=\"131\">$N$<\/span>. The classic Shapley value formula determines this contribution:<\/p>\n<div data-path-to-node=\"21\">\n<div class=\"math-block\" data-math=\"phi_i = sum_{S subseteq N setminus {i}} frac{vert{}Svert{}!(vert{}Nvert{} - vert{}Svert{} - 1)!}{vert{}Nvert{}!} [v(S cup {i}) - v(S)]\">$$phi_i = sum_{S subseteq N setminus {i}} frac{vert{}Svert{}!(vert{}Nvert{} &#8211; vert{}Svert{} &#8211; 1)!}{vert{}Nvert{}!} [v(S cup {i}) &#8211; v(S)]$$<\/div>\n<\/div>\n<p data-path-to-node=\"22\">Where <span class=\"math-inline\" data-math=\"N\" data-index-in-node=\"6\">$N$<\/span> is the universal set of all available features, <span class=\"math-inline\" data-math=\"S\" data-index-in-node=\"56\">$S$<\/span> is a specific subset of features excluding feature <span class=\"math-inline\" data-math=\"i\" data-index-in-node=\"109\">$i$<\/span>, and <span class=\"math-inline\" data-math=\"v(S)\" data-index-in-node=\"116\">$v(S)$<\/span> represents the characteristic function tracking the model&#8217;s output when conditioned only on the features within subset <span class=\"math-inline\" data-math=\"S\" data-index-in-node=\"240\">$S$<\/span>. The expression <span class=\"math-inline\" data-math=\"[v(S cup {i}) - v(S)]\" data-index-in-node=\"258\">$[v(S cup {i}) &#8211; v(S)]$<\/span> isolates the precise marginal shift in credit risk caused by introducing feature <span class=\"math-inline\" data-math=\"i\" data-index-in-node=\"364\">$i$<\/span> to that specific group. By calculating a weighted average of these marginal shifts across every single mathematically viable coalition, SHAP derives a clean, additive attribution value (<span class=\"math-inline\" data-math=\"phi_i\" data-index-in-node=\"552\">$phi_i$<\/span>) for every input.<\/p>\n<p data-path-to-node=\"23\">Unlike LIME, SHAP satisfies critical mathematical axioms that make it incredibly robust for compliance:<\/p>\n<ul data-path-to-node=\"24\">\n<li>\n<p data-path-to-node=\"24,0,0\"><b data-path-to-node=\"24,0,0\" data-index-in-node=\"0\">Local Accuracy (Efficiency):<\/b> The sum of all attribution values matches the difference between the specific model output and the expected baseline output.<\/p>\n<\/li>\n<li>\n<p data-path-to-node=\"24,1,0\"><b data-path-to-node=\"24,1,0\" data-index-in-node=\"0\">Consistency (Monotonicity):<\/b> If a feature&#8217;s marginal impact increases or stays the same when the underlying model changes, its calculated SHAP value cannot decrease. This prevents the computational volatility that impacts LIME.<\/p>\n<\/li>\n<\/ul>\n<h2 data-path-to-node=\"26\"><span class=\"ez-toc-section\" id=\"Technical_Comparison_Evaluating_XAI_Frameworks\"><\/span>Technical Comparison: Evaluating XAI Frameworks<span class=\"ez-toc-section-end\"><\/span><\/h2>\n<p data-path-to-node=\"27\">When designing a production pipeline for automated loan approvals, choosing the right interpretation toolkit involves balancing computational cost with legal reliability.<\/p>\n<table data-path-to-node=\"28\">\n<thead>\n<tr>\n<td><strong>Evaluation Metric<\/strong><\/td>\n<td><strong>LIME (Local Surrogate Approach)<\/strong><\/td>\n<td><strong>SHAP (Shapley Value Approach)<\/strong><\/td>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td><span data-path-to-node=\"28,1,0,0\"><b data-path-to-node=\"28,1,0,0\" data-index-in-node=\"0\">Mathematical Foundation<\/b><\/span><\/td>\n<td><span data-path-to-node=\"28,1,1,0\">Local empirical linear approximations<\/span><\/td>\n<td><span data-path-to-node=\"28,1,2,0\">Cooperative game theory axioms<\/span><\/td>\n<\/tr>\n<tr>\n<td><span data-path-to-node=\"28,2,0,0\"><b data-path-to-node=\"28,2,0,0\" data-index-in-node=\"0\">Computational Overhead<\/b><\/span><\/td>\n<td><span data-path-to-node=\"28,2,1,0\">Low; processes isolated localized samples quickly<\/span><\/td>\n<td><span data-path-to-node=\"28,2,2,0\">High; calculating all feature subsets scales exponentially<\/span><\/td>\n<\/tr>\n<tr>\n<td><span data-path-to-node=\"28,3,0,0\"><b data-path-to-node=\"28,3,0,0\" data-index-in-node=\"0\">Explanation Scope<\/b><\/span><\/td>\n<td><span data-path-to-node=\"28,3,1,0\">Strictly local explanations per applicant<\/span><\/td>\n<td><span data-path-to-node=\"28,3,2,0\">Both local (per-applicant) and global (population-wide) insights<\/span><\/td>\n<\/tr>\n<tr>\n<td><span data-path-to-node=\"28,4,0,0\"><b data-path-to-node=\"28,4,0,0\" data-index-in-node=\"0\">Consistency \/ Stability<\/b><\/span><\/td>\n<td><span data-path-to-node=\"28,4,1,0\">Variable; randomized sampling can yield varying weights<\/span><\/td>\n<td><span data-path-to-node=\"28,4,2,0\">Absolute; identical model states yield identical values<\/span><\/td>\n<\/tr>\n<tr>\n<td><span data-path-to-node=\"28,5,0,0\"><b data-path-to-node=\"28,5,0,0\" data-index-in-node=\"0\">Primary Risk Factor<\/b><\/span><\/td>\n<td><span data-path-to-node=\"28,5,1,0\">Sampling instability can complicate strict audit logs<\/span><\/td>\n<td><span data-path-to-node=\"28,5,2,0\">High feature dimensions require approximations (TreeSHAP)<\/span><\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<h2 data-path-to-node=\"30\"><span class=\"ez-toc-section\" id=\"Bridging_the_Gap_Between_Advanced_Mathematics_and_Industry_Need\"><\/span>Bridging the Gap Between Advanced Mathematics and Industry Need<span class=\"ez-toc-section-end\"><\/span><\/h2>\n<p data-path-to-node=\"31\">Deploying these mathematical models inside real-time financial systems requires a deep understanding of machine learning pipelines, time-series data alignment, and statistical validation. As global regulatory bodies tighten rules around algorithmic fairness, the demand for specialists who can implement explainability frameworks has grown rapidly.<\/p>\n<p data-path-to-node=\"32\">Building models that are highly accurate is no longer enough; data teams must possess the mathematical fluency to unpack those models for non-technical stakeholders, auditors, and regulators. Developing this specialized expertise requires structured education. Professionals aiming to master both predictive modeling and post-hoc explanation frameworks frequently enroll in a comprehensive <a href=\"https:\/\/www.slaconsultantsindia.com\/data-science-r-programming-course.aspx\" target=\"_blank\" rel=\"noopener\"><b data-path-to-node=\"32\" data-index-in-node=\"390\">Data Science course<\/b><\/a> to study the underlying linear algebra, game theory, and structural validation workflows that power modern explainability toolkits.<\/p>\n<p data-path-to-node=\"33\">As major global fintech ecosystems scale up their operations, educational hubs have adapted their training paradigms to address this demand. This evolution is particularly visible in high-growth technology markets, where specialized programs like <a href=\"https:\/\/slaconsultantsdelhi.in\/training-institute-data-science-course\/\" target=\"_blank\" rel=\"noopener\"><b data-path-to-node=\"33\" data-index-in-node=\"247\">Data Science Training in Delhi<\/b><\/a> have heavily updated their curriculums. Today&#8217;s training goes far beyond standard model fitting; it focuses on teaching engineers how to integrate TreeSHAP acceleration into production pipelines, detect hidden algorithmic bias, and build automated compliance engines that meet stringent international regulatory standards.<\/p>\n<h2 data-path-to-node=\"35\"><span class=\"ez-toc-section\" id=\"The_Path_to_Algorithmic_Transparency\"><\/span>The Path to Algorithmic Transparency<span class=\"ez-toc-section-end\"><\/span><\/h2>\n<p data-path-to-node=\"36\">By combining predictive machine learning architectures with SHAP and LIME verification layers, financial institutions no longer have to choose between accuracy and transparency.<\/p>\n<div class=\"code-block ng-tns-c2503008527-40 ng-animate-disabled ng-trigger ng-trigger-codeBlockRevealAnimation\">\n<div class=\"formatted-code-block-internal-container ng-tns-c2503008527-40\">\n<div class=\"animated-opacity ng-tns-c2503008527-40\">\n<pre class=\"ng-tns-c2503008527-40\"><code class=\"code-container formatted ng-tns-c2503008527-40 no-decoration-radius\" role=\"text\" data-test-id=\"code-content\">[Raw Financial Inputs] \u2500\u2500&gt; [High-Accuracy Ensemble Model] \u2500\u2500&gt; [SHAP Evaluation Layer] \u2500\u2500&gt; [Compliant Credit Reason Codes]\n<\/code><\/pre>\n<\/div>\n<\/div>\n<\/div>\n<p data-path-to-node=\"38\">LIME offers quick, flexible glimpses into local decision spaces, while SHAP provides an airtight, mathematically rigorous foundation for systemic compliance auditing. Together, these tools turn complex black-box predictions into clear, auditable insights. They ensure that when an algorithm evaluates an applicant&#8217;s financial future, it does so with a clear rationale that can be explained, audited, and trusted.<\/p>\n<p data-path-to-node=\"39\">How is your engineering team integrating explainability workflows into your automated production models?<\/p>\n","protected":false},"excerpt":{"rendered":"<p>These models are exceptionally good at predicting default risk. But they operate as absolute black boxes.<\/p>\n","protected":false},"author":21001,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[709],"tags":[50470],"class_list":["post-103297","post","type-post","status-publish","format-standard","hentry","category-education","tag-black-box-credit-scoring-models"],"_links":{"self":[{"href":"https:\/\/zamstudios.com\/blogs\/wp-json\/wp\/v2\/posts\/103297","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/zamstudios.com\/blogs\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/zamstudios.com\/blogs\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/zamstudios.com\/blogs\/wp-json\/wp\/v2\/users\/21001"}],"replies":[{"embeddable":true,"href":"https:\/\/zamstudios.com\/blogs\/wp-json\/wp\/v2\/comments?post=103297"}],"version-history":[{"count":1,"href":"https:\/\/zamstudios.com\/blogs\/wp-json\/wp\/v2\/posts\/103297\/revisions"}],"predecessor-version":[{"id":103298,"href":"https:\/\/zamstudios.com\/blogs\/wp-json\/wp\/v2\/posts\/103297\/revisions\/103298"}],"wp:attachment":[{"href":"https:\/\/zamstudios.com\/blogs\/wp-json\/wp\/v2\/media?parent=103297"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/zamstudios.com\/blogs\/wp-json\/wp\/v2\/categories?post=103297"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/zamstudios.com\/blogs\/wp-json\/wp\/v2\/tags?post=103297"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}