xy-scale 1.4.33 → 1.4.35

package/dist/xy-scale.min.js

@@ -1 +1 @@
-var XY_Scale;(()=>{"use strict";var e={d:(r,t)=>{for(var n in t)e.o(t,n)&&!e.o(r,n)&&Object.defineProperty(r,n,{enumerable:!0,get:t[n]})},o:(e,r)=>Object.prototype.hasOwnProperty.call(e,r),r:e=>{"undefined"!=typeof Symbol&&Symbol.toStringTag&&Object.defineProperty(e,Symbol.toStringTag,{value:"Module"}),Object.defineProperty(e,"__esModule",{value:!0})}},r={};e.r(r),e.d(r,{arrayToTimesteps:()=>y,parseProductionX:()=>g,parseTrainingXY:()=>f});const t=e=>null==e||!Number.isFinite(e),n=e=>{const r=[...e];for(let e=r.length-1;e>0;e--){const t=Math.floor(Math.random()*(e+1));[r[e],r[t]]=[r[t],r[e]]}return r},a=(e,{min:r=-1/0,max:t=1/0},n)=>{if(!Array.isArray(e))throw new Error(`Invalid property. "${n}" expected an array.`);if(e.length<r)throw new Error(`Invalid property value. Array "${n}" expected at least ${t} items.`);if(e.length>t)throw new Error(`Invalid property value. Array "${n}" expected at max ${t} items.`);return!0},o=e=>{for(const[r,t]of Object.entries(e)){if("number"==typeof t&&Number.isNaN(t))throw new Error(`Invalid value at index 0 property "${r}": value is "${t}". Expected a numeric value.`);if(null===t)throw new Error(`Invalid value at index 0 property "${r}": value is "${t}".`)}return!0},s=e=>{return Array.isArray(e)?"array":(r=e,"[object Object]"===Object.prototype.toString.call(r)?"object":typeof e);var r},l=e=>Object.keys(e).filter((e=>"tempIdx"!==e)),c=({parentPath:e="",key:r,isArrayParent:t=!1})=>`${e}${t?`[${r}]`:e?`.${r}`:`${r}`}`,i=({rowLabel:e,sourceIndex:r,path:t="",detail:n,strictFlagName:a})=>new Error(`${e} schema error at index "${r}"${t?` path "${t}"`:""}. ${n} Set "${a}" to false to disable this validation.`),u=({referenceRow:e,currentRow:r,rowLabel:t,currentIndex:n,strictFlagName:a})=>{const o=s(e),u=s(r);if(o!==u)throw i({rowLabel:t,sourceIndex:n,detail:`Expected row type "${o}" based on the first parsed ${t} row, but got "${u}".`,strictFlagName:a});if("array"!==o&&"object"!==o)throw i({rowLabel:t,sourceIndex:n,detail:`Parsed ${t} rows must be arrays or plain objects.`,strictFlagName:a});const h=l(e),d=l(r),p=new Set(h),f=new Set(d),g=Array.isArray(e);for(let e=0;e<h.length;e++){const r=h[e];if(!f.has(r))throw i({rowLabel:t,sourceIndex:n,path:c({key:r,isArrayParent:g}),detail:`Missing required ${g?"index":"property"} "${r}" found in the first parsed ${t} row.`,strictFlagName:a})}for(let e=0;e<d.length;e++){const r=d[e];if(!p.has(r))throw i({rowLabel:t,sourceIndex:n,path:c({key:r,isArrayParent:g}),detail:`Unexpected ${g?"index":"property"} "${r}" not present in the first parsed ${t} row.`,strictFlagName:a})}return h},h=e=>Array.isArray(e)?"array":typeof e,d=({referenceValue:e,currentValue:r,sourceIndex:t,path:n,strictFlagName:a})=>{const o=h(e),s=h(r);if(o!==s)throw i({rowLabel:"Y",sourceIndex:t,path:n,detail:`Expected type "${o}" based on the first parsed Y row, but got "${s}".`,strictFlagName:a});if("array"!==o){if("number"!==o&&"boolean"!==o&&"string"!==o)throw i({rowLabel:"Y",sourceIndex:t,path:n,detail:`Unsupported Y value type "${o}". Y values must be numbers, booleans, strings, or nested arrays of those types.`,strictFlagName:a})}else{if(e.length!==r.length)throw i({rowLabel:"Y",sourceIndex:t,path:n,detail:`Expected array length "${e.length}" based on the first parsed Y row, but got "${r.length}".`,strictFlagName:a});for(let o=0;o<e.length;o++)d({referenceValue:e[o],currentValue:r[o],sourceIndex:t,path:c({parentPath:n,key:o,isArrayParent:!0}),strictFlagName:a})}},p=({rows:e,sourceIndexes:r,strictFlagName:t})=>{if(!e.length)return;const n=e[0],a=(e,r)=>{u({referenceRow:n,currentRow:e,rowLabel:"X",currentIndex:r,strictFlagName:t})};e.length>1&&a(e[e.length-1],r[e.length-1]);for(let t=0;t<e.length;t++)a(e[t],r[t])},f=({arrObj:e=[],trainingSplit:r=.8,yCallbackFunc:s=e=>e,xCallbackFunc:i=e=>e,validateRows:h=()=>!0,shuffle:f=!1,balancing:g="",strictXSchema:y=!0,strictYSchema:w=!0,state:b={}})=>{let m=[],x=[];const $=[];a(e,{min:5},"parseTrainingXY"),o(e[0]);for(let r=0;r<e.length;r++){if(!h({objRow:e,index:r,state:b}))continue;const t=i({objRow:e,index:r,state:b}),n=s({objRow:e,index:r,state:b});null!=t&&null!=n&&(m.push(t),x.push(n),$.push(r))}if(y&&p({rows:m,sourceIndexes:$,strictFlagName:"strictXSchema"}),w&&(({rows:e,sourceIndexes:r,strictFlagName:t})=>{if(!e.length)return;const n=e[0],a=(e,r)=>{const a=u({referenceRow:n,currentRow:e,rowLabel:"Y",currentIndex:r,strictFlagName:t}),o=Array.isArray(n);for(let s=0;s<a.length;s++){const l=a[s],i=c({key:l,isArrayParent:o});d({referenceValue:n[l],currentValue:e[l],sourceIndex:r,path:i,strictFlagName:t})}};e.length>1&&a(e[e.length-1],r[e.length-1]);for(let t=0;t<e.length;t++)a(e[t],r[t])})({rows:x,sourceIndexes:$,strictFlagName:"strictYSchema"}),f){const e=new Array(m.length);for(let r=0;r<m.length;r++)e[r]={x:m[r],y:x[r],sourceIndex:$[r]};const r=n(e);m=new Array(r.length),x=new Array(r.length);for(let e=0;e<r.length;e++)m[e]=r[e].x,x[e]=r[e].y,$[e]=r[e].sourceIndex}const v=m.length,I=x.length,A=v?l(m[0]):[],N=I?l(x[0]):[],F=new Array(v),j=new Array(I),Y={keyNames:A},X={keyNames:N};for(let e=0;e<v;e++){const r=m[e],n=$[e],a=new Array(A.length);for(let e=0;e<A.length;e++){const o=A[e],s=r[o];if(t(s))throw new Error(`Invalid property value (${s}) returned from "xCallbackFunc" on index "${n}" property "${o}".`);a[e]=s}F[e]=a}for(let e=0;e<I;e++){const r=x[e],t=new Array(N.length);for(let e=0;e<N.length;e++)t[e]=r[N[e]];j[e]=t}const S=Math.floor(F.length*r);let E=F.slice(0,S),k=j.slice(0,S),O=F.slice(S),P=j.slice(S);if(g){let e;if("oversample"===g)e=((e,r)=>{const t={},n={};r.forEach(((a,o)=>{t[a]||(t[a]=0,n[a]=[]),t[a]++,n[a].push([e[o],r[o]])}));const a=Math.max(...Object.values(t)),o=[],s=[];return Object.keys(n).forEach((e=>{const r=n[e],t=r.length;for(let e=0;e<a;e++){const n=r[e%t];o.push(n[0]),s.push(n[1])}})),{X:o,Y:s}})(E,k),E=e.X,k=e.Y;else{if("undersample"!==g)throw Error('balancing argument only accepts "", "oversample" and "undersample". Defaults to "".');e=((e,r)=>{const t={},n={};r.forEach(((a,o)=>{t[a]||(t[a]=0,n[a]=[]),t[a]++,n[a].push([e[o],r[o]])}));const a=Math.min(...Object.values(t)),o=[],s=[];return Object.keys(n).forEach((e=>{const r=n[e];for(let e=0;e<a;e++){const t=r[e];o.push(t[0]),s.push(t[1])}})),{X:o,Y:s}})(E,k),E=e.X,k=e.Y}}return{trainX:E,trainY:k,testX:O,testY:P,configX:Y,configY:X}},g=({arrObj:e=[],xCallbackFunc:r=e=>e,validateRows:s=()=>!0,shuffle:c=!1,strictXSchema:i=!0,state:u={}})=>{let h=[],d=[];a(e,{min:5},"parseProductionX"),o(e[0]);for(let t=0;t<e.length;t++){if(!s({objRow:e,index:t,state:u}))continue;const n=r({objRow:e,index:t,state:u});null!=n&&!1!==n&&(h.push(n),d.push(t))}if(i&&p({rows:h,sourceIndexes:d,strictFlagName:"strictXSchema"}),c){const e=new Array(h.length);for(let r=0;r<h.length;r++)e[r]={x:h[r],sourceIndex:d[r]};const r=n(e);h=new Array(r.length),d=new Array(r.length);for(let e=0;e<r.length;e++)h[e]=r[e].x,d[e]=r[e].sourceIndex}const f=h.length,g=f?l(h[0]):[],y=new Array(f),w={keyNames:g};for(let e=0;e<f;e++){const r=h[e],n=d[e],a=new Array(g.length);for(let e=0;e<g.length;e++){const o=g[e],s=r[o];if(t(s))throw new Error(`Invalid property value (${s}) returned from "xCallbackFunc" on index "${n}" property "${o}".`);a[e]=s}y[e]=a}return{X:y,configX:w}},y=(e,r)=>{if(0===r)return e;if(r<0)throw new Error("timeSteps must be greater than 0");const t=[];for(let n=0;n<=e.length-r;n++)t.push(e.slice(n,n+r));return t};XY_Scale=r})();
+var XY_Scale;(()=>{"use strict";var e={d:(r,t)=>{for(var n in t)e.o(t,n)&&!e.o(r,n)&&Object.defineProperty(r,n,{enumerable:!0,get:t[n]})},o:(e,r)=>Object.prototype.hasOwnProperty.call(e,r),r:e=>{"undefined"!=typeof Symbol&&Symbol.toStringTag&&Object.defineProperty(e,Symbol.toStringTag,{value:"Module"}),Object.defineProperty(e,"__esModule",{value:!0})}},r={};e.r(r),e.d(r,{arrayToTimesteps:()=>A,parseProductionX:()=>v,parseTrainingXY:()=>$});const t=e=>null==e||!Number.isFinite(e),n=e=>{const r=[...e];for(let e=r.length-1;e>0;e--){const t=Math.floor(Math.random()*(e+1));[r[e],r[t]]=[r[t],r[e]]}return r},a=e=>"string"==typeof e||"boolean"==typeof e||("number"==typeof e?Number.isFinite(e):!!Array.isArray(e)&&e.every(a)),o=({random:e,seed:r}={})=>"function"==typeof e?e:Number.isInteger(r)?(e=>{let r=e>>>0;return()=>{r+=1831565813;let e=Math.imul(r^r>>>15,1|r);return e^=e+Math.imul(e^e>>>7,61|e),((e^e>>>14)>>>0)/4294967296}})(r):Math.random,s=(e,r=Math.random)=>{for(let t=e.length-1;t>0;t--){const n=Math.floor(r()*(t+1));[e[t],e[n]]=[e[n],e[t]]}return e},l=(e,r,t=Math.random)=>{if(r>e.length)throw new Error("Cannot sample more items than available without replacement.");const n=[...e];return s(n,t),n.slice(0,r)},i=(e,r,t=Math.random)=>{if(0===e.length)throw new Error("Cannot sample from an empty array.");const n=[];for(let a=0;a<r;a++){const r=Math.floor(t()*e.length);n.push(e[r])}return n},c=(e,r)=>r?structuredClone(e):e,h=(e,r,{cloneX:t=!1}={})=>{((e,r)=>{if(!Array.isArray(e)||!Array.isArray(r))throw new Error("X and Y must be arrays.");if(e.length!==r.length)throw new Error("X and Y must have the same length.");if(0===e.length)throw new Error("X and Y cannot be empty.")})(e,r);const n=new Map;return r.forEach(((r,o)=>{const s=(e=>{if(!a(e))throw new Error("Invalid Y label. Allowed types: finite numbers, strings, booleans, or nested arrays of those.");return JSON.stringify(e)})(r);n.has(s)||n.set(s,[]),n.get(s).push({x:c(e[o],t),y:r})})),n},u=(e,{min:r=-1/0,max:t=1/0},n)=>{if(!Array.isArray(e))throw new Error(`Invalid property. "${n}" expected an array.`);if(e.length<r)throw new Error(`Invalid property value. Array "${n}" expected at least ${t} items.`);if(e.length>t)throw new Error(`Invalid property value. Array "${n}" expected at max ${t} items.`);return!0},d=e=>{for(const[r,t]of Object.entries(e)){if("number"==typeof t&&Number.isNaN(t))throw new Error(`Invalid value at index 0 property "${r}": value is "${t}". Expected a numeric value.`);if(null===t)throw new Error(`Invalid value at index 0 property "${r}": value is "${t}".`)}return!0},f=e=>{return Array.isArray(e)?"array":(r=e,"[object Object]"===Object.prototype.toString.call(r)?"object":typeof e);var r},p=e=>Object.keys(e).filter((e=>"tempIdx"!==e)),y=({parentPath:e="",key:r,isArrayParent:t=!1})=>`${e}${t?`[${r}]`:e?`.${r}`:`${r}`}`,g=({rowLabel:e,sourceIndex:r,path:t="",detail:n,strictFlagName:a})=>new Error(`${e} schema error at index "${r}"${t?` path "${t}"`:""}. ${n} Set "${a}" to false to disable this validation.`),m=({referenceRow:e,currentRow:r,rowLabel:t,currentIndex:n,strictFlagName:a})=>{const o=f(e),s=f(r);if(o!==s)throw g({rowLabel:t,sourceIndex:n,detail:`Expected row type "${o}" based on the first parsed ${t} row, but got "${s}".`,strictFlagName:a});if("array"!==o&&"object"!==o)throw g({rowLabel:t,sourceIndex:n,detail:`Parsed ${t} rows must be arrays or plain objects.`,strictFlagName:a});const l=p(e),i=p(r),c=new Set(l),h=new Set(i),u=Array.isArray(e);for(let e=0;e<l.length;e++){const r=l[e];if(!h.has(r))throw g({rowLabel:t,sourceIndex:n,path:y({key:r,isArrayParent:u}),detail:`Missing required ${u?"index":"property"} "${r}" found in the first parsed ${t} row.`,strictFlagName:a})}for(let e=0;e<i.length;e++){const r=i[e];if(!c.has(r))throw g({rowLabel:t,sourceIndex:n,path:y({key:r,isArrayParent:u}),detail:`Unexpected ${u?"index":"property"} "${r}" not present in the first parsed ${t} row.`,strictFlagName:a})}return l},w=e=>Array.isArray(e)?"array":typeof e,b=({referenceValue:e,currentValue:r,sourceIndex:t,path:n,strictFlagName:a})=>{const o=w(e),s=w(r);if(o!==s)throw g({rowLabel:"Y",sourceIndex:t,path:n,detail:`Expected type "${o}" based on the first parsed Y row, but got "${s}".`,strictFlagName:a});if("array"!==o){if("number"!==o&&"boolean"!==o&&"string"!==o)throw g({rowLabel:"Y",sourceIndex:t,path:n,detail:`Unsupported Y value type "${o}". Y values must be numbers, booleans, strings, or nested arrays of those types.`,strictFlagName:a})}else{if(e.length!==r.length)throw g({rowLabel:"Y",sourceIndex:t,path:n,detail:`Expected array length "${e.length}" based on the first parsed Y row, but got "${r.length}".`,strictFlagName:a});for(let o=0;o<e.length;o++)b({referenceValue:e[o],currentValue:r[o],sourceIndex:t,path:y({parentPath:n,key:o,isArrayParent:!0}),strictFlagName:a})}},x=({rows:e,sourceIndexes:r,strictFlagName:t})=>{if(!e.length)return;const n=e[0],a=(e,r)=>{m({referenceRow:n,currentRow:e,rowLabel:"X",currentIndex:r,strictFlagName:t})};e.length>1&&a(e[e.length-1],r[e.length-1]);for(let t=0;t<e.length;t++)a(e[t],r[t])},$=({arrObj:e=[],trainingSplit:r=.8,yCallbackFunc:a=e=>e,xCallbackFunc:f=e=>e,validateRows:g=()=>!0,shuffle:w=!1,balancing:$="",strictXSchema:v=!0,strictYSchema:A=!0,state:I={}})=>{let N=[],X=[];const F=[];u(e,{min:5},"parseTrainingXY"),d(e[0]);for(let r=0;r<e.length;r++){if(!g({objRow:e,index:r,state:I}))continue;const t=f({objRow:e,index:r,state:I}),n=a({objRow:e,index:r,state:I});null!=t&&null!=n&&(N.push(t),X.push(n),F.push(r))}if(v&&x({rows:N,sourceIndexes:F,strictFlagName:"strictXSchema"}),A&&(({rows:e,sourceIndexes:r,strictFlagName:t})=>{if(!e.length)return;const n=e[0],a=(e,r)=>{const a=m({referenceRow:n,currentRow:e,rowLabel:"Y",currentIndex:r,strictFlagName:t}),o=Array.isArray(n);for(let s=0;s<a.length;s++){const l=a[s],i=y({key:l,isArrayParent:o});b({referenceValue:n[l],currentValue:e[l],sourceIndex:r,path:i,strictFlagName:t})}};e.length>1&&a(e[e.length-1],r[e.length-1]);for(let t=0;t<e.length;t++)a(e[t],r[t])})({rows:X,sourceIndexes:F,strictFlagName:"strictYSchema"}),w){const e=new Array(N.length);for(let r=0;r<N.length;r++)e[r]={x:N[r],y:X[r],sourceIndex:F[r]};const r=n(e);N=new Array(r.length),X=new Array(r.length);for(let e=0;e<r.length;e++)N[e]=r[e].x,X[e]=r[e].y,F[e]=r[e].sourceIndex}const Y=N.length,E=X.length,S=Y?p(N[0]):[],j=E?p(X[0]):[],M=new Array(Y),R=new Array(E),k={keyNames:S},P={keyNames:j};for(let e=0;e<Y;e++){const r=N[e],n=F[e],a=new Array(S.length);for(let e=0;e<S.length;e++){const o=S[e],s=r[o];if(t(s))throw new Error(`Invalid property value (${s}) returned from "xCallbackFunc" on index "${n}" property "${o}".`);a[e]=s}M[e]=a}for(let e=0;e<E;e++){const r=X[e],t=new Array(j.length);for(let e=0;e<j.length;e++)t[e]=r[j[e]];R[e]=t}const O=Math.floor(M.length*r);let L=M.slice(0,O),C=R.slice(0,O),T=M.slice(O),V=R.slice(O);if($){let e;if("oversample"===$)e=((e,r,t={})=>{const{random:n,seed:a,shuffleResult:l=!0,cloneX:u=!1}=t,d=o({random:n,seed:a}),f=h(e,r,{cloneX:u}),p=[...f.values()].map((e=>e.length)),y=Math.max(...p),g=[];for(const e of f.values()){const r=[...e],t=y-r.length,n=t>0?i(e,t,d).map((e=>({x:c(e.x,u),y:e.y}))):[];g.push(...r,...n)}return l&&s(g,d),{X:g.map((({x:e})=>e)),Y:g.map((({y:e})=>e))}})(L,C),L=e.X,C=e.Y;else{if("undersample"!==$)throw Error('balancing argument only accepts "", "oversample" and "undersample". Defaults to "".');e=((e,r,t={})=>{const{random:n,seed:a,shuffleResult:i=!0,cloneX:u=!1}=t,d=o({random:n,seed:a}),f=h(e,r,{cloneX:u}),p=[...f.values()].map((e=>e.length)),y=Math.min(...p),g=[];for(const e of f.values()){const r=l(e,y,d).map((e=>({x:c(e.x,u),y:e.y})));g.push(...r)}return i&&s(g,d),{X:g.map((({x:e})=>e)),Y:g.map((({y:e})=>e))}})(L,C),L=e.X,C=e.Y}}return{trainX:L,trainY:C,testX:T,testY:V,configX:k,configY:P}},v=({arrObj:e=[],xCallbackFunc:r=e=>e,validateRows:a=()=>!0,shuffle:o=!1,strictXSchema:s=!0,state:l={}})=>{let i=[],c=[];u(e,{min:5},"parseProductionX"),d(e[0]);for(let t=0;t<e.length;t++){if(!a({objRow:e,index:t,state:l}))continue;const n=r({objRow:e,index:t,state:l});null!=n&&!1!==n&&(i.push(n),c.push(t))}if(s&&x({rows:i,sourceIndexes:c,strictFlagName:"strictXSchema"}),o){const e=new Array(i.length);for(let r=0;r<i.length;r++)e[r]={x:i[r],sourceIndex:c[r]};const r=n(e);i=new Array(r.length),c=new Array(r.length);for(let e=0;e<r.length;e++)i[e]=r[e].x,c[e]=r[e].sourceIndex}const h=i.length,f=h?p(i[0]):[],y=new Array(h),g={keyNames:f};for(let e=0;e<h;e++){const r=i[e],n=c[e],a=new Array(f.length);for(let e=0;e<f.length;e++){const o=f[e],s=r[o];if(t(s))throw new Error(`Invalid property value (${s}) returned from "xCallbackFunc" on index "${n}" property "${o}".`);a[e]=s}y[e]=a}return{X:y,configX:g}},A=(e,r,t=1)=>{if(!Array.isArray(e))throw new Error("arr must be an array");if(!Number.isInteger(r)||r<=0)throw new Error("timeSteps must be a positive integer");if(!Number.isInteger(t)||t<=0)throw new Error("step must be a positive integer");if(r>e.length)return[];const n=[];for(let a=0;a<=e.length-r;a+=t)n.push(e.slice(a,a+r));return n};XY_Scale=r})();

package/index.js

@@ -1,4 +1,5 @@
 import { parseTrainingXY, parseProductionX } from "./src/datasets.js"
 import {arrayToTimesteps } from "./src/timeSteps.js"
+import { arrayShuffle, arrayShuffleXY } from "./src/utilities.js"
 
 export { parseTrainingXY, parseProductionX, arrayToTimesteps }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "xy-scale",
-  "version": "1.4.33",
+  "version": "1.4.35",
   "main": "./index.js",
   "type": "module",
   "scripts": {

package/src/balancing.js

@@ -1,71 +1,187 @@
-export const oversampleXY = (X, Y) => {
-
-    const labelCounts = {};
-    const data = {};
-  
-    // Count occurrences of each label and group by label
-    Y.forEach((label, i) => {
-      if (!labelCounts[label]) {
-        labelCounts[label] = 0;
-        data[label] = [];
-      }
-      labelCounts[label]++;
-      data[label].push([X[i], Y[i]]);
-    });
-  
-    // Find the max label count
-    const maxCount = Math.max(...Object.values(labelCounts));
-  
-    const oversampledX = [];
-    const oversampledY = [];
-  
-    // Oversample each label to match the max count
-    Object.keys(data).forEach(label => {
-      const samples = data[label];
-      const numSamples = samples.length;
-  
-      for (let i = 0; i < maxCount; i++) {
-        const sample = samples[i % numSamples]; // Cycle through samples
-        oversampledX.push(sample[0]);
-        oversampledY.push(sample[1]);
-      }
-    });
-  
-    return { X: oversampledX, Y: oversampledY };
+const isValidLabelValue = (value) => {
+  if (typeof value === "string" || typeof value === "boolean") return true;
+  if (typeof value === "number") return Number.isFinite(value);
+
+  if (Array.isArray(value)) {
+    return value.every(isValidLabelValue);
   }
-  
-
-export const undersampleXY = (X, Y) => {
-
-    const labelCounts = {};
-    const data = {};
-  
-    // Count occurrences of each label and group by label
-    Y.forEach((label, i) => {
-      if (!labelCounts[label]) {
-        labelCounts[label] = 0;
-        data[label] = [];
-      }
-      labelCounts[label]++;
-      data[label].push([X[i], Y[i]]);
-    });
-  
-    // Find the minimum label count
-    const minCount = Math.min(...Object.values(labelCounts));
-  
-    const undersampledX = [];
-    const undersampledY = [];
-  
-    // Undersample each label to match the minimum count
-    Object.keys(data).forEach(label => {
-      const samples = data[label];
-      for (let i = 0; i < minCount; i++) {
-        const sample = samples[i]; // Use first minCount samples
-        undersampledX.push(sample[0]);
-        undersampledY.push(sample[1]);
-      }
+
+  return false;
+};
+
+const makeLabelKey = (label) => {
+  if (!isValidLabelValue(label)) {
+    throw new Error(
+      "Invalid Y label. Allowed types: finite numbers, strings, booleans, or nested arrays of those."
+    );
+  }
+
+  return JSON.stringify(label);
+};
+
+const validateXY = (X, Y) => {
+  if (!Array.isArray(X) || !Array.isArray(Y)) {
+    throw new Error("X and Y must be arrays.");
+  }
+
+  if (X.length !== Y.length) {
+    throw new Error("X and Y must have the same length.");
+  }
+
+  if (X.length === 0) {
+    throw new Error("X and Y cannot be empty.");
+  }
+};
+
+const mulberry32 = (seed) => {
+  let t = seed >>> 0;
+
+  return () => {
+    t += 0x6D2B79F5;
+    let r = Math.imul(t ^ (t >>> 15), t | 1);
+    r ^= r + Math.imul(r ^ (r >>> 7), r | 61);
+    return ((r ^ (r >>> 14)) >>> 0) / 4294967296;
+  };
+};
+
+const resolveRandom = ({ random, seed } = {}) => {
+  if (typeof random === "function") return random;
+  if (Number.isInteger(seed)) return mulberry32(seed);
+  return Math.random;
+};
+
+const shuffleInPlace = (arr, random = Math.random) => {
+  for (let i = arr.length - 1; i > 0; i--) {
+    const j = Math.floor(random() * (i + 1));
+    [arr[i], arr[j]] = [arr[j], arr[i]];
+  }
+
+  return arr;
+};
+
+const sampleWithoutReplacement = (samples, size, random = Math.random) => {
+  if (size > samples.length) {
+    throw new Error("Cannot sample more items than available without replacement.");
+  }
+
+  const copy = [...samples];
+  shuffleInPlace(copy, random);
+  return copy.slice(0, size);
+};
+
+const sampleWithReplacement = (samples, size, random = Math.random) => {
+  if (samples.length === 0) {
+    throw new Error("Cannot sample from an empty array.");
+  }
+
+  const out = [];
+
+  for (let i = 0; i < size; i++) {
+    const idx = Math.floor(random() * samples.length);
+    out.push(samples[idx]);
+  }
+
+  return out;
+};
+
+const maybeClone = (value, clone) => {
+  if (!clone) return value;
+  return structuredClone(value);
+};
+
+const groupXYByLabel = (X, Y, { cloneX = false } = {}) => {
+  validateXY(X, Y);
+
+  const groups = new Map();
+
+  Y.forEach((label, i) => {
+    const key = makeLabelKey(label);
+
+    if (!groups.has(key)) {
+      groups.set(key, []);
+    }
+
+    groups.get(key).push({
+      x: maybeClone(X[i], cloneX),
+      y: label,
     });
-  
-    return { X: undersampledX, Y: undersampledY };
+  });
+
+  return groups;
+};
+
+export const oversampleXY = (X, Y, options = {}) => {
+  const {
+    random,
+    seed,
+    shuffleResult = true,
+    cloneX = false,
+  } = options;
+
+  const rng = resolveRandom({ random, seed });
+  const groups = groupXYByLabel(X, Y, { cloneX });
+
+  const counts = [...groups.values()].map((samples) => samples.length);
+  const maxCount = Math.max(...counts);
+
+  const combined = [];
+
+  for (const samples of groups.values()) {
+    const originals = [...samples];
+    const extrasNeeded = maxCount - originals.length;
+
+    const extras =
+      extrasNeeded > 0
+        ? sampleWithReplacement(samples, extrasNeeded, rng).map((sample) => ({
+            x: maybeClone(sample.x, cloneX),
+            y: sample.y,
+          }))
+        : [];
+
+    combined.push(...originals, ...extras);
+  }
+
+  if (shuffleResult) {
+    shuffleInPlace(combined, rng);
+  }
+
+  return {
+    X: combined.map(({ x }) => x),
+    Y: combined.map(({ y }) => y),
+  };
+};
+
+export const undersampleXY = (X, Y, options = {}) => {
+  const {
+    random,
+    seed,
+    shuffleResult = true,
+    cloneX = false,
+  } = options;
+
+  const rng = resolveRandom({ random, seed });
+  const groups = groupXYByLabel(X, Y, { cloneX });
+
+  const counts = [...groups.values()].map((samples) => samples.length);
+  const minCount = Math.min(...counts);
+
+  const combined = [];
+
+  for (const samples of groups.values()) {
+    const selected = sampleWithoutReplacement(samples, minCount, rng).map((sample) => ({
+      x: maybeClone(sample.x, cloneX),
+      y: sample.y,
+    }));
+
+    combined.push(...selected);
   }
-  
+
+  if (shuffleResult) {
+    shuffleInPlace(combined, rng);
+  }
+
+  return {
+    X: combined.map(({ x }) => x),
+    Y: combined.map(({ y }) => y),
+  };
+};

package/src/timeSteps.js

@@ -1,12 +1,23 @@
-export const arrayToTimesteps = (arr, timeSteps) => {
-    if (timeSteps === 0) return arr;
-    if (timeSteps < 0) throw new Error("timeSteps must be greater than 0");
-    
-    const timestepsArray = [];
-    
-    for (let i = 0; i <= arr.length - timeSteps; i++) {
-        timestepsArray.push(arr.slice(i, i + timeSteps));
-    }
-    
-    return timestepsArray;
-}
+export const arrayToTimesteps = (arr, timeSteps, step = 1) => {
+  // Validation
+  if (!Array.isArray(arr)) {
+    throw new Error('arr must be an array');
+  }
+  if (!Number.isInteger(timeSteps) || timeSteps <= 0) {
+    throw new Error('timeSteps must be a positive integer');
+  }
+  if (!Number.isInteger(step) || step <= 0) {
+    throw new Error('step must be a positive integer');
+  }
+
+  // If no windows possible, return empty array
+  if (timeSteps > arr.length) {
+    return [];
+  }
+
+  const timestepsArray = [];
+  for (let i = 0; i <= arr.length - timeSteps; i += step) {
+    timestepsArray.push(arr.slice(i, i + timeSteps));
+  }
+  return timestepsArray;
+};

package/src/utilities.js

@@ -11,4 +11,28 @@ export const arrayShuffle = X => {
     }
 
     return shuffledX;
-}
+}
+
+export const arrayShuffleXY = (X, Y) => {
+    if (!Array.isArray(X) || !Array.isArray(Y)) {
+        throw new TypeError('Both X and Y must be arrays');
+    }
+
+    if (X.length !== Y.length) {
+        throw new Error('X and Y must have the same length');
+    }
+
+    // Copy to avoid mutating the originals
+    const shuffledX = [...X];
+    const shuffledY = [...Y];
+
+    // Shuffle both arrays using the same swaps
+    for (let i = shuffledX.length - 1; i > 0; i--) {
+        const j = Math.floor(Math.random() * (i + 1));
+
+        [shuffledX[i], shuffledX[j]] = [shuffledX[j], shuffledX[i]];
+        [shuffledY[i], shuffledY[j]] = [shuffledY[j], shuffledY[i]];
+    }
+
+    return { X: shuffledX, Y: shuffledY };
+};

package/test/test.js

@@ -1,5 +1,6 @@
 import OHLCV_INDICATORS from 'ohlcv-indicators'
 import { parseTrainingXY } from "../src/datasets.js"
+import { arrayToTimesteps } from '../src/timeSteps.js'
 import { loadFile } from "./fs.js"
 
 const test = async () => {
@@ -60,6 +61,12 @@ const test = async () => {
     console.log('row_1', {features: trainX[0], labels: trainY[0]})
 
     console.log(trainY.length, trainX.length)
+
+    const timeSteps = arrayToTimesteps(trainX, 10)
+
+    const typeArr = (a) => Array.isArray(a) ? 'array' : typeof a
+
+    console.log(`timeSteps: ${typeArr(timeSteps)} => ${typeArr(timeSteps[0])} => ${typeArr(timeSteps[0][0][0])}`)
 }
 
 //callback function used to prepare X before flattening